Intelligent scheduling method and system for multi-process combined regeneration of waste sand
By constructing a data-driven closed-loop control system and optimizing the waste sand recycling production line using a virtual mapping model and dynamic scheduling algorithm, the problems of poor inter-process coordination and rigid scheduling were solved, achieving efficient and stable waste sand recycling treatment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ANHUI UNIVERSITY OF ARCHITECTURE
- Filing Date
- 2026-04-13
- Publication Date
- 2026-05-12
AI Technical Summary
In existing waste sand recycling production lines, the control systems of each process are independent of each other, lacking global information sharing and coordination mechanisms. This leads to material congestion in one stage while equipment in another stage is idle. The rigid scheduling strategy cannot adapt to changes in raw material characteristics and lacks foresight, resulting in poor processing efficiency and energy waste.
By constructing a data-driven closed-loop control system, real-time operating condition datasets of multiple processes are obtained. Virtual mapping models are used to predict the dynamic correlation between processes. Dynamic scheduling instructions are generated by combining scheduling algorithms. The production process is optimized through real-time feedback and dynamic adjustment mechanisms, achieving deep collaboration and dynamic optimization of the entire process.
It improves the continuity of the production line and the overall processing efficiency, enhances the adaptability to fluctuations in raw material characteristics and equipment failures, ensures the smooth operation of the production process and the stability of the quality of recycled products, and has the ability to self-evolve to continuously improve efficiency.
Smart Images
Figure CN122022402A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of waste sand recycling and intelligent control technology, and relates to an intelligent scheduling method and system for multi-process joint recycling of waste sand. Background Technology
[0002] Waste sand recycling, especially in industries such as foundry, refers to the process of treating used sand through a series of physical or chemical methods to restore its usability for recycling. This process typically involves multiple closely linked steps, including crushing, screening, magnetic separation, washing, drying, and cooling. The efficiency and coordination of these steps directly determine the quality of the recycled sand, the recycling rate, energy consumption, and the final production cost. Therefore, efficient production scheduling across these multiple steps is crucial.
[0003] In existing waste sand recycling production lines, scheduling mainly relies on manual experience or simple, rule-based automated control systems. Operators determine the start-up and shutdown times and operating parameters of each process's equipment based on on-site observation and past experience. In some highly automated production lines, although programmable logic controllers (PLCs) are used to control individual devices, the control systems for each process typically operate independently. The connection between processes and material flow mainly rely on preset, static logical rules, such as triggering the start of downstream equipment through level gauge signals.
[0004] However, existing technologies have obvious technical defects: First, the control systems of each process are independent of each other and lack a global information sharing and coordination mechanism, resulting in poor coordination between processes and the phenomenon that materials are congested in one link while equipment in another link is idle.
[0005] Secondly, the rigid scheduling strategy based on fixed rules cannot adapt to the characteristics of waste sand raw materials, such as dynamic changes in humidity and impurity content, often resulting in poor treatment effect or unnecessary energy waste.
[0006] Furthermore, existing scheduling methods lack foresight regarding the production process, and can only respond passively to potential bottlenecks or equipment malfunctions, resulting in delayed responses and making it difficult to optimize overall production efficiency. Summary of the Invention
[0007] In order to overcome the above-mentioned defects of the prior art and to achieve the above objectives, the present invention proposes the following technical solution: The first embodiment of the present invention provides an intelligent scheduling method for multi-process joint regeneration of waste sand, including: acquiring the waste sand characteristic parameters and equipment status parameters of multiple processes in the waste sand regeneration process, performing fusion processing, and generating a real-time operating condition dataset.
[0008] The real-time operating condition dataset is input into a preset virtual mapping model to perform impact prediction and output process collaboration prediction data, wherein the virtual mapping model is used to characterize the dynamic relationship between processes.
[0009] Based on the process collaborative prediction data, and combined with a scheduling algorithm for balancing multi-dimensional production goals, an initial scheduling scheme is generated.
[0010] Based on the initial scheduling scheme and by introducing dynamically changing equipment constraints, dynamic scheduling instructions are generated.
[0011] The waste sand recycling equipment is controlled to execute the dynamic scheduling command, and feedback data during the execution process is collected and used to update the real-time operating condition dataset.
[0012] The second embodiment of the present invention provides an intelligent scheduling system for multi-process joint regeneration of waste sand, including: a data fusion module, a predictive analysis module, a scheme formulation module, a dynamic scheduling module, and a control execution module.
[0013] The data fusion module is connected to the predictive analysis module, the predictive analysis module is connected to the scheme formulation module, the scheme formulation module is connected to the dynamic scheduling module, the dynamic scheduling module is connected to the control execution module, and the control execution module is connected to the data fusion module.
[0014] The data fusion module is used to acquire waste sand characteristic parameters and equipment status parameters, and combine them with historical operating data to generate a real-time operating data set.
[0015] The predictive analysis module is used to process the real-time working condition dataset through a preset virtual mapping model and output process collaborative prediction data.
[0016] The scheme formulation module is used to generate an initial scheduling scheme based on the process collaborative prediction data and a scheduling algorithm for balancing multi-dimensional production goals.
[0017] The dynamic scheduling module is used to generate dynamic scheduling instructions based on the initial scheduling scheme and by introducing equipment constraints.
[0018] The control execution module is used to control the waste sand recycling equipment to execute the dynamic scheduling instructions, collect feedback data and transmit it to the data fusion module to update the real-time operating condition dataset.
[0019] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0020] (1) This invention achieves deep collaboration and dynamic optimization of the entire process of waste sand recycling by constructing a data-driven closed-loop control, integrating each independent process into an organic whole, and predicting the chain reaction between processes through a virtual mapping model, so that scheduling decisions are no longer isolated, but based on global optimal considerations, thereby reducing bottlenecks and waiting between processes and improving the process continuity and comprehensive processing efficiency of the entire production line.
[0021] (2) By introducing a real-time feedback and dynamic adjustment mechanism, especially a flexible elastic time window, the present invention enables the scheduling scheme to enhance its adaptability to uncertain factors such as fluctuations in raw material characteristics and minor equipment failures, and improves the adaptability and robustness of the production system to external disturbances and internal changes, thereby ensuring the smooth operation of the production process and the long-term stability of the quality of recycled products.
[0022] (3) This invention endows the scheduling system with the ability to continuously learn and self-evolve. Through regular scheduling effect evaluation and iterative optimization of models and algorithms, it continuously improves its own decision-making logic. This self-evolution mechanism helps to maintain the long-term applicability and effectiveness of the scheduling strategy, enabling the entire production system to continuously improve with the changes in time and environment, and achieve continuous improvement in efficiency. Attached Figure Description
[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0024] Figure 1 A flowchart illustrating the implementation steps of the method provided in the first embodiment of the present invention.
[0025] Figure 2 This is a schematic diagram of the dynamic scheduling instruction generation process in the first embodiment of the present invention.
[0026] Figure 3 This is a schematic diagram of module connections provided for the second embodiment of the present invention. Detailed Implementation
[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0028] Please see Figure 1As shown, the first embodiment of the present invention provides an intelligent scheduling method for multi-process joint regeneration of waste sand, including: S1. Obtaining the waste sand characteristic parameters and equipment status parameters of multiple processes in the waste sand regeneration process, performing fusion processing, and generating a real-time operating condition dataset.
[0029] In a specific embodiment of the present invention, the step of generating a real-time operating condition dataset includes:
[0030] Real-time parameters from multiple processes during waste sand recycling are collected and preprocessed to obtain raw state data.
[0031] It should be noted that the real-time parameter collection for multiple processes in the waste sand recycling process involves the following:
[0032] Industrial-grade sensors are deployed at key process nodes in waste sand recycling to collect real-time parameters reflecting the characteristics of waste sand and the operating status of equipment. These key process nodes can be exemplified as the feed inlet and outlet of the crusher, screening equipment, washing tank, and drying kiln. The sensors continuously collect information during operation, forming a multi-dimensional raw sensor data stream containing timestamps.
[0033] For the collection of waste sand characteristic parameters, near-infrared humidity sensors can be used to obtain the moisture content data of waste sand, or laser particle size analyzers can be used to monitor the particle size distribution of crushed particles.
[0034] For collecting equipment status parameters, current transformers can be installed on the equipment motors to obtain their operating load information, or vibration acceleration sensors can be installed to monitor the mechanical health of the equipment.
[0035] The preprocessing of real-time parameters for the above multiple processes involves the following steps:
[0036] Digital filtering techniques are used to smooth the data in order to filter out high-frequency noise and extract effective trend information. The digital filtering technique can be exemplarily a moving average filter, which creates a series of averages by calculating the average of consecutive subsets in the data sequence, thereby effectively smoothing short-term fluctuations.
[0037] Convert real-time parameters from multiple processes into a standardized data structure, such as... or The format generates structured data containing parameter names, values, units, and acquisition timestamps, thereby eliminating the problem of inconsistent data formats output by sensors from different manufacturers and of different types.
[0038] The real-time parameters of multiple processes, converted to a standard data structure, are standardized. Specifically, the min-max normalization method is used to linearly map the values of each dimension to... Within the interval, the calculation formula is: ,in These are the standardized values obtained after calculation. This represents the actual value of a certain dimension parameter obtained from the preprocessed data. and These are the minimum and maximum normal range values for this dimension parameter as defined in historical operating data or equipment manuals.
[0039] By retrieving historical data related to the current operating condition from a database containing pre-stored historical production information, and based on multi-dimensional similarity matching between the current operating condition characteristics and historical operating condition records, historical data related to the current operating condition is obtained.
[0040] It should be noted that the specific process for retrieving historical data related to the current operating conditions is as follows:
[0041] The raw state data is constructed into a current feature vector containing multiple key process parameters according to a preset parameter list. This vector represents the instantaneous operating state of the current process.
[0042] The historical database stores numerous sets of historical feature vectors with the same dimension as the current feature vector in historical preset production cycles, as well as the final production results corresponding to each set. The production results include, but are not limited to, unit energy consumption, recycled sand quality grade, and total processing time.
[0043] The similarity distance between the current feature vector and each historical feature vector in the historical database is calculated in a multidimensional space. This process uses the Euclidean distance algorithm for quantitative measurement. The smaller the Euclidean distance value obtained, the higher the similarity between the historical working condition record and the current working condition in the multidimensional feature space.
[0044] The historical feature vectors in the historical database are sorted in ascending order of Euclidean distance. A preset number of historical feature vectors and their final production results are then selected as historical data related to the current working conditions.
[0045] The original state data and the historical operating condition data are weighted and fused to generate the real-time operating condition dataset, wherein the fusion weights are dynamically allocated based on the real-time confidence and historical similarity of the data sources.
[0046] It should be added that the dynamic allocation of the fusion weights follows the principle of real-time confidence-driven and historical similarity-weighted, and the weighted fusion process can be exemplarily referred to the formula. In this formula, This represents the new value of a certain feature parameter generated after fusion, and this value will be used as a component of the multi-source fused data. It is the real-time measurement value of this parameter in the current feature vector. Is this parameter in the first... The values in a highly similar historical feature vector. This refers to the numbering of each highly similar historical feature vector. , It is the first The weights of highly similar historical feature vectors are proportional to the similarity between the case and the current state, and the sum of all weights is 1, ensuring that historical data with higher similarity accounts for a larger proportion in the fusion process. It is a preset confidence factor, with a value between 0 and 1, used to adjust the relative importance of real-time data and historical data in the fusion result. When the device is running stably or the sensor data has high reliability, i.e., the real-time confidence of the data source is high, The value can be increased to emphasize real-time performance, or decreased to rely more on historical experience.
[0047] By performing this weighted fusion calculation on each parameter in the current feature vector, a multi-source fusion data that reflects both the current real operating conditions and contains historical operating patterns is generated.
[0048] S2. Input the real-time operating condition dataset into a preset virtual mapping model to perform impact prediction and output process collaboration prediction data, wherein the virtual mapping model is used to characterize the dynamic relationship between processes.
[0049] In a specific embodiment of the present invention, the construction and application of the virtual mapping model includes:
[0050] An initial relational model is obtained by training based on historical production information using machine learning methods.
[0051] It should be noted that the above training process using machine learning methods includes:
[0052] From historical production information, a production record dataset containing multiple batches and the entire process is extracted. Each record contains three types of features:
[0053] a) Input status parameters, such as the characteristics of the material when entering a certain process.
[0054] b) Process operation parameters, such as equipment settings for the process.
[0055] c) Output result parameters, such as the material characteristics after the completion of the process, and the energy consumption and efficiency of the process.
[0056] Before inputting the model, the cleaning, standardization, and labeling of three types of features need to be completed.
[0057] To capture the complex nonlinear and temporal dependencies between process parameters, advanced machine learning algorithms such as gradient boosting decision trees or long short-term memory networks are selected for model training.
[0058] During training, the output parameters of the upstream process and the operating parameters of the current process are used as input features, and the key performance indicators of the current process or the characteristic parameters of the output materials are used as prediction targets.
[0059] By iteratively calculating on a large amount of data and using mean squared error as the loss function, the error between the predicted value and the true value is continuously reduced. The model eventually learns a complex mapping function from input features to output target.
[0060] After training, the model's performance is evaluated using a reserved test dataset to ensure that its prediction accuracy meets application requirements. This yields an initial relational model capable of predicting process results based on given input.
[0061] Based on the correlation analysis between historical process parameters, the dependence strength between processes is quantified, and process chain weight factors are generated.
[0062] The quantification process of inter-process dependency strength can employ existing SHAP analysis methods. The specific quantification logic is as follows: For any downstream process's performance indicator, all potentially affected upstream process output parameters are analyzed. For each affected upstream process output parameter, a weight coefficient is determined. This weight coefficient reflects the magnitude and importance of the change in the downstream process's performance indicator caused by a unit change in the upstream process's output parameter. A larger absolute value of the weight coefficient indicates a higher dependence of the upstream process on that downstream process's performance indicator, i.e., a tighter coupling relationship. Through this process, corresponding process chain weight factors are calculated for all key inter-process relationships.
[0063] The process chain weight factors are integrated into the initial relationship model, and the virtual mapping model is formed by adjusting the contribution of the upstream process output to the downstream process input.
[0064] It should be added that the process chain weight factor is not used to modify the model's input data, but rather serves as a quantitative indicator of the strength of inter-process correlation. It is stored in the knowledge base and plays a role in subsequent scheduling decision-making stages. Specifically, after using the virtual mapping model to perform impact prediction and output process collaboration prediction data, the process chain weight factor will be used as an important reference, inputting into the scheduling algorithm of the scheme formulation module. This is used to quantify and adjust the comprehensive impact weight of changes in a certain process parameter on the overall production target, thereby enabling the generated initial scheduling scheme to more reasonably balance the coupling relationships between various processes.
[0065] This invention, through the construction of a data-driven closed-loop control, achieves deep collaboration and dynamic optimization of the entire waste sand recycling process, integrating each independent process into an organic whole. By predicting the chain reactions between processes through a virtual mapping model, scheduling decisions are no longer isolated but based on global optimal considerations, thereby reducing bottlenecks and waiting between processes and improving the process continuity and overall processing efficiency of the entire production line.
[0066] S3. Based on the process collaborative prediction data, and combined with the scheduling algorithm used to balance multi-dimensional production goals, an initial scheduling scheme is generated.
[0067] In a specific embodiment of the present invention, the step of generating an initial scheduling scheme includes:
[0068] The process collaboration prediction data is analyzed, and a process priority list is generated by quantifying the comprehensive impact of each process on multi-dimensional production objectives. The multi-dimensional production objectives include energy consumption, quality, and efficiency of production tasks.
[0069] It should be noted that the process collaborative prediction data is a data structure containing predicted values for multiple key performance indicators, such as the predicted total energy consumption for the next hour, the final qualified rate of recycled sand, and the estimated processing time for each process. To quantify the comprehensive impact of each process on multi-dimensional production targets, a priority scoring function can be used for calculation. In this formula, Representing the Process priority score, A higher score indicates a greater deviation of the current process from the overall target, and therefore requires priority adjustment. The numbers representing the various performance evaluation dimensions. Such as energy consumption, quality, efficiency, etc. These are preset weighting coefficients, reflecting the weighting of the first [unit / group] under the current production target. The importance of each performance evaluation dimension is determined by process engineers based on the production plan; these weights are set accordingly. The results of the process impact prediction regarding the first... The process is in the first step Predicted values across each performance evaluation dimension It is the first Preset target values or optimal benchmark values for each performance evaluation dimension.
[0070] Based on the process priority list and combined with equipment operation constraints, optimized settings for equipment operation parameters are generated through multi-objective optimization calculations to form the initial scheduling scheme.
[0071] It should be noted that the multi-objective optimization calculation logic starts from the highest priority process and uses the particle swarm optimization algorithm to find a set of optimal equipment operating parameters. The optimization objective is to minimize the priority score of the process, and the optimization boundary is strictly limited by preset equipment constraints. These constraints include the physical limits of the equipment and process specification requirements. Physical limits include the maximum power of the motor and the maximum safe temperature of the drying kiln, while process specification requirements include the maximum residence time of materials in a certain process. The specific implementation of the particle swarm optimization algorithm is as follows:
[0072] The multiple equipment operation parameters to be optimized in the aforementioned process are constructed into a multi-dimensional search space, where the upper and lower bounds of each dimension are strictly limited by the equipment constraints. Each potential solution in the search space is encoded as a particle, and the position vector of each particle represents a specific combination of equipment operation parameters. At the same time, a velocity vector is initialized for each particle. It is used to control the direction and step size of parameter search.
[0073] Within the parameter range defined by the device constraints, an initial population containing N particles is randomly generated, where N is the preset population size, and the initial velocity of each particle is set to a random value within the preset range.
[0074] The fitness function is defined as the priority score calculation function of the process. For the position vector of each particle, it is substituted into the preset virtual mapping model to predict the process performance index under the set of operation parameters, and then its corresponding priority score is calculated. The optimization objective of the particle swarm optimization algorithm is to minimize the fitness function value.
[0075] In each iteration, the particle state is updated according to the following rules:
[0076] (1) Evaluate the fitness value of each particle's current position and update the individual's historical best position. and global optimal position .
[0077] (2) Update the velocity and position of each particle in the D-dimensional dimension according to the following formula:
[0078]
[0079]
[0080] in, For inertial weights, and For acceleration coefficient, and A random number in the range [0,1]. and These represent the velocity and position of particle z in the D-th dimension, respectively. The number representing each particle. .
[0081] (3) Apply boundary constraints to the updated particle positions to ensure that they are always within the feasible domain defined by the device constraints.
[0082] When the preset termination condition is met, the combination of equipment operation parameters corresponding to the globally optimal position is output as the optimal operation setting for the process.
[0083] S4. Based on the initial scheduling scheme and by introducing dynamically changing equipment constraints, generate dynamic scheduling instructions.
[0084] Please see Figure 2 As shown, in a specific embodiment of the present invention, the step of generating dynamic scheduling instructions includes:
[0085] Based on the buffer capacity between processes and the historical execution time fluctuation range, the time margin of each process is calculated, and elasticity parameters are generated.
[0086] It should be noted that the process of calculating the time margin of each process and generating the flexibility parameters is described step by step as follows:
[0087] First, a series of basic parameters are collected and established for each process to be calculated. These basic parameters include: the standard theoretical execution time of the process obtained from the process manual and equipment performance data; the historical average execution time obtained from long-term operation data statistics; the statistical standard deviation used to quantify the historical completion time fluctuation of the process; and the maximum buffer time that the process can be allowed, which is jointly determined by the capacity of upstream and downstream equipment and the requirements for process continuity.
[0088] The time margin is calculated based on the aforementioned fundamental parameters. The logic is to select the smaller value between the maximum allowable buffer time and the historical execution time fluctuation range as the initial candidate value for the time margin. The historical execution time fluctuation range is determined by multiplying a preset elasticity coefficient by the standard deviation of the historical execution time of the process. This preset elasticity coefficient is used to adjust the system's tolerance to normal time fluctuations, and its value is typically within a certain range. To ensure the practical significance of the time margin, the calculated result is compared with zero, and the final determined time margin is the larger value, thus ensuring that it is always a non-negative number.
[0089] The calculated time margin is combined with the standard theoretical execution time of the process to form a flexible parameter that characterizes the time characteristics of the process. This parameter clearly indicates the theoretically required execution time of the process, as well as the time buffer margin that can be flexibly used in actual production.
[0090] The above steps are repeated for all key processes in the production process to generate unique elastic parameters for each process. Finally, these parameters are aggregated to form a complete set of elastic parameters covering the entire production line. This set provides a crucial data foundation for developing production scheduling schemes that are resistant to interference and adaptable.
[0091] The elastic parameters are applied to the initial scheduling scheme, expanding the fixed time nodes into flexible execution time windows, and generating adjusted scheduling instructions.
[0092] The adjusted scheduling instructions undergo conflict detection. If a conflict is detected, the elastic parameters and priorities of relevant processes are adjusted collaboratively to eliminate the conflict. The adjusted instructions are then iteratively verified. During the iterative verification process, a maximum iteration threshold is set. If the number of iterations exceeds this threshold and unresolved conflicts still exist, it is determined that a conflict-free ideal scheduling instruction cannot be generated within the current scheduling cycle. In this case, a degradation processing strategy is initiated: based on the rule of prioritizing the highest priority process, a scheduling scheme with the lowest conflict level is forcibly executed, and conflict information is recorded for subsequent optimization of the elastic parameter generation rules. Simultaneously, an alarm is triggered, prompting manual intervention. Conversely, if consensus is reached within the threshold, the final dynamic scheduling instruction is output.
[0093] In a specific embodiment of the present invention, the generation of the elastic parameter further includes:
[0094] Based on the prediction results of future operating condition fluctuations in the process collaborative prediction data, the value of the elasticity parameter is dynamically adjusted to generate an adaptive elasticity parameter.
[0095] It should be noted that the technical details of the above-mentioned adaptive elasticity parameter generation are as follows:
[0096] From the process co-prediction data, volatility predictions for key operating conditions within a specific future production cycle are extracted. These results are typically quantified into one or more specific volatility indicators, such as: the predicted rate of change in raw material impurity content, the efficiency degradation trend of key equipment, or the expected intensity of the impact of ambient temperature and humidity on the drying process.
[0097] A pre-defined adjustment rule library containing multiple "condition-action" rules is provided. Each rule defines how to adjust the elasticity parameters of the corresponding process when a specific future fluctuation in operating conditions is predicted. For example:
[0098] Example of rule A: If it is predicted that the average hardness of the waste sand entering the crushing process will rise above the threshold within a preset time period in the future, the time margin in the elasticity parameter of the crushing process will be increased by a preset percentage based on the original calculated value.
[0099] Example of rule B: If it is predicted that the heat exchange efficiency of the drying kiln will decrease beyond the threshold due to ash accumulation in the future cycle, then the time margin of the drying process will be increased by a fixed buffer amount based on the original value.
[0100] The acquired predictive volatility indicators are matched with the adjustment rule base. For rules that match successfully, the original elasticity parameters of the relevant processes are calculated and corrected according to the adjustment logic defined by the rules. The resulting new parameters are the adaptive elasticity parameters, which retain the statistical rationality based on historical data and incorporate predictive compensation for future volatility.
[0101] The adjusted scheduling instructions are generated using the adaptive elasticity parameters, thereby enabling the scheduling instructions to adapt to future production changes in a predictable manner.
[0102] The embodiments of the present invention introduce a real-time feedback and dynamic adjustment mechanism, especially a flexible and elastic time window, which enables the scheduling scheme to enhance its adaptability to uncertainties such as fluctuations in raw material characteristics and minor equipment failures. This improves the adaptability and robustness of the production system to external disturbances and internal changes, thereby ensuring the smooth operation of the production process and the long-term stability of the quality of recycled products.
[0103] S5. Control the waste sand recycling equipment to execute the dynamic scheduling command, collect feedback data during the execution process, and update the real-time operating condition dataset accordingly.
[0104] In a specific embodiment of the present invention, the application of the feedback data includes:
[0105] The feedback data during the execution process is compared with the expected execution target in the dynamic scheduling instruction to calculate the execution deviation.
[0106] It should be noted that the execution deviation can be defined as the absolute difference between the actual value fed back by the sensor and the expected execution target value in the instruction, and then the result of the ratio calculation with the expected execution target value.
[0107] The execution deviation is fed back to the virtual mapping model, triggering online parameter correction of the model.
[0108] It should be noted that the execution deviation is fed back to the virtual mapping model as follows: the execution deviation and its corresponding operating condition data are used as training samples to trigger online parameter correction of the virtual mapping model. Specifically, the stochastic gradient descent method is used to reduce the mean square error between the model prediction and the actual observation. The model weights are fine-tuned through the backpropagation algorithm so that the model can dynamically adapt to changes in equipment performance and fluctuations in raw material characteristics.
[0109] By updating the process coordination prediction data using the revised virtual mapping model, closed-loop optimization of scheduling instructions can be achieved.
[0110] In a specific embodiment of the present invention, the method further includes: performing trend analysis on the execution deviation within a continuous scheduling cycle, identifying systematic performance drift, and generating a strategy adjustment factor.
[0111] It should be noted that the trend analysis of the execution deviation within the continuous scheduling cycle can use the root mean square error to quantify the overall deviation between the actual state and the expected state. When the root mean square error shows a continuous upward trend, for example, a 10% increase in the number of consecutive preset scheduling cycles, it is determined that there is a systematic performance drift.
[0112] The strategy adjustment factor is generated based on the drift feature, and its value range is [0.5, 1.5]. In a specific embodiment, the strategy adjustment factor can be linearly calculated based on the rate of change of the root mean square error. Specifically, the preset adjustment coefficient is multiplied by the rate of change of the root mean square error, and the sum of the product and 1 is used as the strategy adjustment factor to ensure that the strategy adjustment factor is positively correlated with the rate of change of the root mean square error.
[0113] Based on the aforementioned strategy adjustment factor, the weight allocation of multi-dimensional production objectives in the scheduling algorithm is dynamically adjusted.
[0114] It should be noted that the weight allocation of multi-dimensional production objectives in the scheduling algorithm is dynamically adjusted as follows: the energy consumption weight is updated to the product of the strategy adjustment factor and the original weight; the efficiency weight is updated to the product of the reciprocal of the strategy adjustment factor and the original weight; and the quality weight is updated to the product of 2, the difference between the strategy adjustment factor and the original weight. It should also be noted that after completing the above adjustments, all updated weights need to be normalized to ensure that the sum of the weights of all objectives is 1, maintaining the balance of the optimization objectives.
[0115] An initial scheduling scheme is generated using a scheduling algorithm that updates weights, enabling the autonomous evolution and continuous optimization of the scheduling strategy.
[0116] In a specific embodiment of the present invention, the method further includes: dynamically adjusting the confidence weights of the original state data and historical operating condition data during the fusion process based on the statistical characteristics of the execution deviation.
[0117] It should be noted that the confidence weights of the original state data and historical operating condition data are dynamically adjusted during the fusion process as follows: the sliding standard deviation of the recent deviation is calculated, and the reciprocal of the sum of the sliding standard deviation and 1 is used as the confidence weight of the original state data. When the system stability decreases, the reference weight of historical data is increased, and when the system is running stably, the focus is on real-time data. Through this adaptive mechanism, it is ensured that the real-time operating condition dataset always reflects the optimal and reliable state of the system.
[0118] The original state data and historical operating condition data are weighted and fused using the dynamic confidence weight to generate a real-time operating condition dataset, enabling the data fusion process to have adaptive capability to the system's credibility.
[0119] This invention endows scheduling with the ability to continuously learn and self-evolve. Through regular evaluation of scheduling effectiveness and iterative optimization of models and algorithms, it continuously improves its decision-making logic. This self-evolution mechanism helps maintain the long-term applicability and effectiveness of scheduling strategies, enabling the entire production system to continuously improve with changes in time and environment, and achieve continuous improvement in efficiency.
[0120] Please see Figure 3 As shown, the second embodiment of the present invention provides an intelligent scheduling system for multi-process joint regeneration of waste sand, including: a data fusion module, a predictive analysis module, a scheme formulation module, a dynamic scheduling module, and a control execution module.
[0121] The data fusion module is connected to the predictive analysis module, the predictive analysis module is connected to the scheme formulation module, the scheme formulation module is connected to the dynamic scheduling module, the dynamic scheduling module is connected to the control execution module, and the control execution module is connected to the data fusion module.
[0122] The data fusion module is used to acquire waste sand characteristic parameters and equipment status parameters, and combine them with historical operating data to generate a real-time operating data set.
[0123] The predictive analysis module is used to process the real-time working condition dataset through a preset virtual mapping model and output process collaborative prediction data.
[0124] The scheme formulation module is used to generate an initial scheduling scheme based on the process collaborative prediction data and a scheduling algorithm for balancing multi-dimensional production goals.
[0125] The dynamic scheduling module is used to generate dynamic scheduling instructions based on the initial scheduling scheme and by introducing equipment constraints.
[0126] The control execution module is used to control the waste sand recycling equipment to execute the dynamic scheduling instructions, collect feedback data and transmit it to the data fusion module to update the real-time operating condition dataset.
[0127] It should be noted that the formulas described above, through the principles of dimensional consistency and mathematical standardization, can translate physical quantities with different properties into unitless standard values or superimposed parameters of the same dimension. This eliminates the interference of different dimensions on the computational logic, allowing the formulas to retain the original data distribution characteristics while possessing mathematical rationality and adaptability to objective laws. The descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the invention.
[0128] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.
Claims
1. An intelligent scheduling method for multi-process combined recycling of waste sand, characterized in that, Includes the following steps: The waste sand characteristic parameters and equipment status parameters of multiple processes in the waste sand regeneration process are obtained, fused, and used to generate a real-time operating condition dataset. The real-time operating condition dataset is input into a preset virtual mapping model to perform impact prediction and output process collaboration prediction data, wherein the virtual mapping model is used to characterize the dynamic relationship between processes. Based on the process collaborative prediction data, and combined with the scheduling algorithm used to balance multi-dimensional production goals, an initial scheduling scheme is generated. Based on the initial scheduling scheme, dynamically changing equipment constraints are introduced to generate dynamic scheduling instructions. The waste sand recycling equipment is controlled to execute the dynamic scheduling command, and feedback data during the execution process is collected and used to update the real-time operating condition dataset.
2. The intelligent scheduling method for multi-process joint regeneration of waste sand according to claim 1, characterized in that, The steps for generating the real-time operating condition dataset include: Real-time parameters of multiple processes in the waste sand recycling process are collected and preprocessed to obtain the original state data; From a database containing pre-stored historical production information, based on multi-dimensional similarity matching between current operating condition characteristics and historical operating condition records, historical data related to the current operating condition is retrieved to obtain historical operating condition data; The original state data and the historical operating condition data are weighted and fused to generate the real-time operating condition dataset, wherein the fusion weights are dynamically allocated based on the real-time confidence and historical similarity of the data sources.
3. The intelligent scheduling method for multi-process joint regeneration of waste sand according to claim 2, characterized in that, The construction and application of the virtual mapping model include: An initial relationship model is obtained by training based on historical production information using machine learning methods. Based on the correlation analysis between historical process parameters, the dependence strength between processes is quantified, and process chain weight factors are generated. The process chain weight factors are integrated into the initial relationship model, and the virtual mapping model is formed by adjusting the contribution of the upstream process output to the downstream process input.
4. The intelligent scheduling method for multi-process joint regeneration of waste sand according to claim 1, characterized in that, The steps for generating the initial scheduling scheme include: The process collaboration prediction data is analyzed, and a process priority list is generated by quantifying the comprehensive impact of each process on multi-dimensional production objectives. The multi-dimensional production objectives include the energy consumption, quality, and efficiency of production tasks. Based on the process priority list and combined with equipment operation constraints, optimized settings for equipment operation parameters are generated through multi-objective optimization calculations to form the initial scheduling scheme. The multi-objective optimization calculation adopts the particle swarm optimization algorithm, and the parameters of each process are optimized in order of process priority from high to low.
5. The intelligent scheduling method for multi-process joint regeneration of waste sand according to claim 4, characterized in that, The step of generating dynamic scheduling instructions includes: Based on the buffer capacity between processes and the historical execution time fluctuation range, the time margin of each process is calculated, and elasticity parameters are generated. The elastic parameters are applied to the initial scheduling scheme to expand the fixed time nodes into flexible execution time windows, generating adjusted scheduling instructions. The adjusted scheduling instructions are subjected to conflict detection. If a conflict is detected, the elastic parameters and priorities of the relevant processes are adjusted in a coordinated manner to eliminate the conflict. The adjusted instructions are then iterated and verified a limited number of times until the elastic execution windows of all processes are coordinated and conflict-free, and the final dynamic scheduling instructions are output.
6. The intelligent scheduling method for multi-process joint regeneration of waste sand according to claim 5, characterized in that, The generation of the elastic parameters also includes: Based on the prediction results of future operating condition fluctuations in the process collaboration prediction data, the value of the elasticity parameter is dynamically adjusted to generate an adaptive elasticity parameter. The adjusted scheduling instructions are generated using the adaptive resilience parameters, thereby enabling the scheduling instructions to adapt to future production changes in a predictable manner.
7. The intelligent scheduling method for multi-process joint regeneration of waste sand according to claim 1, characterized in that, The applications of the feedback data include: The feedback data during the execution process is compared with the expected execution target in the dynamic scheduling instruction to calculate the execution deviation; The execution deviation is fed back to the virtual mapping model, triggering online parameter correction of the model; By updating the process coordination prediction data using the revised virtual mapping model, closed-loop optimization of scheduling instructions can be achieved.
8. The intelligent scheduling method for multi-process joint regeneration of waste sand according to claim 1, characterized in that, Also includes: Perform trend analysis on execution deviations within continuous scheduling cycles to identify systemic performance drift and generate strategy adjustment factors; Based on the aforementioned strategy adjustment factor, the weight allocation of multi-dimensional production objectives in the scheduling algorithm is dynamically adjusted. An initial scheduling scheme is generated using a scheduling algorithm that updates weights, enabling the autonomous evolution and continuous optimization of the scheduling strategy.
9. The intelligent scheduling method for multi-process joint regeneration of waste sand according to claim 1, characterized in that, Also includes: Based on the statistical characteristics of the execution deviation, the confidence weights of the original state data and historical operating condition data are dynamically adjusted during the fusion process. The original state data and historical operating condition data are weighted and fused using the confidence weights to generate a real-time operating condition dataset, enabling the data fusion process to adapt to the system's confidence level.
10. An intelligent scheduling system for multi-process combined recycling of waste sand, characterized in that, include: The data fusion module is used to acquire waste sand characteristic parameters and equipment status parameters, and combine them with historical operating data to generate a real-time operating data set. The predictive analysis module is used to process the real-time working condition dataset through a virtual mapping model and output process collaborative prediction data. The scheme formulation module is used to generate an initial scheduling scheme based on the process collaborative prediction data and a scheduling algorithm for balancing multi-dimensional production goals. The dynamic scheduling module is used to generate dynamic scheduling instructions based on the initial scheduling scheme and by introducing equipment constraints. The control execution module is used to control the waste sand recycling equipment to execute the dynamic scheduling instructions, collect feedback data and transmit it to the data fusion module to update the real-time operating condition dataset.