An intelligent production scheduling method for a foundry machining production line

Abstract

The application discloses an intelligent production scheduling method of a foundry processing production line, and particularly relates to the technical field of production line intelligent production scheduling, and through time sequence collection of multiple source production elements and energy consumption-temperature-output correlation mapping, a scheduling feasible space capable of dynamically sensing thermal field changes is constructed; on the basis, a multi-objective weight self-adaptive drift mechanism is introduced, so that the scheduling algorithm can self-adjust the weight according to the actual load and energy consumption gradient, and continuous self-correction of the optimization direction of the beat and energy consumption is realized; further, in combination with optimization direction phase offset evaluation and scheduling shock risk index calculation, active identification and inhibition capacity of the chain instability phenomenon of 'energy consumption lag-weight drift-direction shock' is formed, so that the optimization process still maintains convergence and stability under dynamic conditions, and through embedding the risk index into the scheduling iteration process, a penalty correction is implemented on the optimization direction.

Description

Technical Field

[0001] This invention relates to the field of landscape lighting control technology, and more specifically, to an intelligent production scheduling method for a casting production line. Background Technology

[0002] Traditional casting production lines typically include multiple processes such as molding, pouring, cooling, sand removal, cleaning, and machining. These processes are subject to strong constraints in their sequence and energy consumption timing. In the intelligent scheduling of casting production lines, there is a significant coupling relationship between the dynamic characteristics of the thermal field and the multi-objective optimization weight adaptive mechanism. When thermal field lag exists in the production system, energy consumption feedback signals often lag behind actual temperature changes, causing the energy consumption monitoring module to deviate in the instantaneous state. When the scheduling model performs real-time optimization based on this, it incorrectly assumes that the system's thermal load has decreased, thus adjusting the processing cycle or equipment start-up and shutdown strategies, leading to a mismatch between energy consumption estimation and the actual thermal state. Simultaneously, the multi-objective weight adaptive module in the scheduling algorithm (dynamically balancing between shortest production time and lowest energy consumption) updates weights based on this "false signal," causing the optimization direction to frequently switch between different objectives. When weights frequently drift, the energy consumption prediction and cycle control of the scheduling system exhibit a mutually reinforcing positive feedback effect, forming a cyclical chain of "thermal field lag - energy consumption misestimation - weight drift - scheduling switch - thermal field lag again." This loop disrupts the convergence and stability of the algorithm, leading to problems such as oscillations in production scheduling optimization direction, unstable energy consumption control, and mismatched production cycle time. Essentially, this is a self-excited optimization oscillation phenomenon caused by the coupling of energy consumption feedback lag and adaptive weight update mechanism, which is a key technical bottleneck restricting the accuracy and stability of intelligent scheduling in casting production lines. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide an intelligent production scheduling method for a casting processing production line to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A method for intelligent scheduling of casting production lines includes the following steps: Collect sensor data on equipment status, mold temperature, pouring cycle and energy consumption during the casting process to form a multi-source production factor sequence with timestamps; The multi-source production factor sequence is processed by feature denoising to extract parameters such as production cycle fluctuation, thermal field gradient and energy consumption distribution, and an energy consumption-temperature-output correlation mapping model is established. A dynamic feasibility map is constructed based on the association mapping model, which transforms the processing capacity, heat load and energy consumption feedback of each equipment station into node states, forming a scheduling feasibility space with time-varying weights. A multi-objective weight adaptive drift mechanism is introduced in the scheduling feasible space. By monitoring real-time production load, equipment health and energy consumption gradient, the objective function weight vector is dynamically adjusted to form a self-correcting optimization direction constraint structure. The optimization direction constraint structure is input into the nonlinear optimizer to establish the optimization direction phase offset evaluation function. By calculating the time delay coupling between the target gradient change and energy consumption feedback, the potential triggering conditions for scheduling direction oscillation are identified. By comprehensively weighting the thermal field energy consumption feedback lag information, the multi-objective weight adaptive drift characteristics, and the optimization direction phase offset index, the scheduling oscillation risk index is calculated, thereby realizing the dynamic capture of unstable scheduling trends. By embedding the scheduling oscillation risk index into the real-time scheduling iteration process, the optimization direction is penalized and corrected.

[0005] In a preferred embodiment, the multi-source production factor sequence is subjected to feature denoising processing to extract parameters such as production cycle fluctuation, thermal gradient, and energy consumption distribution, and an energy consumption-temperature-output correlation mapping model is established, as follows: Perform sliding window mean filtering for noise reduction on the aligned signal of each channel: ,in The sampled signal after denoising. To adjust the sliding window size, The time index within the sliding window; The extraction of production cycle fluctuations includes based on the pouring cycle signal. Extracting beat intervals : And calculate the volatility indicator: ,in The total number of samples, The average beat interval, For the amplitude of rhythm fluctuation The extraction of the thermal field gradient includes based on the mold temperature array Calculate the thermal gradient: ,in , These represent the rate of temperature change along the x and y directions of the mold plane, respectively. For thermal field gradient; The extraction of energy consumption distribution parameters includes the analysis of energy consumption signals. Perform time series statistics and sliding window analysis: , ,in This represents the average energy consumption. For energy consumption fluctuations, For the energy consumption analysis window size, This serves as the time index within the energy consumption analysis window. The establishment of the energy consumption-temperature-output correlation mapping model includes inputting the cycle time fluctuation amplitude, thermal field gradient, and energy consumption fluctuation into a multivariate regression model: ,in For output quality indicators, For regression coefficients, For residual terms; The output quality index is compared with the preset output quality threshold. If the output quality index is greater than the output quality threshold, a dynamic feasibility map needs to be constructed based on the correlation mapping model to provide constraints and target references for production scheduling optimization. Combine cycle time fluctuation amplitude, thermal gradient, energy consumption fluctuation, and output quality indicators into a unified time series vector: .

[0006] In a preferred embodiment, a dynamic feasibility map is constructed based on an association mapping model, transforming the processing capacity, thermal load, and energy consumption feedback of each equipment station into node states, forming a scheduling feasibility space with time-varying weights, as detailed below: The processing capacity of each equipment station is determined by the cycle time fluctuation. The heat load of each equipment station is measured by the thermal field gradient. The energy consumption feedback for each equipment station is measured by energy consumption fluctuations. measure; Define the process path dependency matrix: ; For any two adjacent device nodes u and v, define the energy consumption coupling weight: ,in For energy consumption coupling weights, For the energy consumption fluctuation of device node u For the energy consumption fluctuation of device node v Let be the thermal gradient of device node u. Let be the thermal gradient of device node v. , This is a preset proportionality coefficient between the energy consumption difference and the heat load difference, and , All are greater than 0; Integrate all device nodes and energy consumption coupling weights into a dynamic graph structure: ,in For a set of device nodes, For the set of process-connected edges, The set of energy consumption coupling weights is used to calculate the comprehensive feasibility score for each device node. ,in The scheduling feasibility score for the device node at the current moment is given. Let u be the cycle time fluctuation amplitude of device node u. These are preset proportional coefficients for the cycle time fluctuation amplitude, energy consumption fluctuation, and thermal gradient of device node u, respectively. All are greater than 0; A time-varying feasible space matrix is ​​constructed based on the comprehensive feasibility score and energy consumption coupling weight. : ,in It is an energy consumption suppression and regulation factor.

[0007] In a preferred embodiment, a multi-objective weight adaptive drift mechanism is introduced into the scheduling feasible space. By monitoring real-time production load, equipment health, and energy consumption gradient, the objective function weight vector is dynamically adjusted to form a self-correcting optimization direction constraint structure, as detailed below: Based on the production scheduling objective, establish an initial weight vector: ,in These are the initial weights for processing time, energy consumption, and equipment load balancing objectives, respectively. The equipment load rate at the current moment is collected through the production monitoring system. Equipment health With energy consumption gradient Calculate the dynamic feedback factor : ,in This represents the historical maximum value of the equipment load rate. This represents the historical maximum value of the energy consumption gradient. This is an empirical adjustment coefficient; Based on dynamic feedback factors Update weight vector : ; Based on the updated weight vector, the direction of scheduling evaluation optimization is redefined in the scheduling feasible space: ,in This is the output value of the comprehensive scheduling evaluation function; Through real-time iteration The gradient changes are used to identify the drift trend of the objective function and form an adaptive optimization direction constraint structure. ,in for The gradient change.

[0008] In a preferred embodiment, the optimization direction constraint structure is input into a nonlinear optimizer to establish an optimization direction phase offset evaluation function. By calculating the time delay coupling between the target gradient change and energy consumption feedback, potential triggering conditions for scheduling direction oscillations are identified, as follows: The specific formula for the optimized direction phase offset evaluation function is as follows: ,in Let be the objective function. Represents the scheduling decision variables; During the optimization iteration process, the gradient change of the objective function at adjacent time points is recorded. : ,in For time step; The gradient change sequence is transformed into a complex signal using the Hilbert transform, which is then used to extract the phase of the target gradient change. ,in For Hilbert transform operators, The imaginary unit, The signal is a complex analytic signal consisting of the gradient change and the imaginary part of the Hilbert transform. The phase of the target gradient change; Based on the timing difference between equipment energy consumption and scheduling response, the energy consumption feedback delay phase of the energy consumption feedback signal is defined. : ,in The dominant frequency component of energy consumption variation. Energy consumption feedback time delay; By calculating the time delay coupling between the target gradient change and energy consumption feedback. The details are as follows: ; The time-delay coupling amount is compared with the preset time-delay coupling amount threshold. If the time-delay coupling amount is greater than the time-delay coupling amount threshold, it is determined that the scheduling direction oscillation has been triggered.

[0009] In a preferred embodiment, if a scheduling direction oscillation is triggered, the scheduling oscillation risk index is calculated by comprehensively weighting the thermal field energy consumption feedback lag information, the multi-objective weight adaptive drift characteristics, and the optimized direction phase offset index. The thermal field energy consumption feedback lag information is measured by the thermal field energy consumption lag coefficient, and the calculation formula for the thermal field energy consumption lag coefficient is as follows: ,in The thermal field energy consumption lag coefficient, To calculate the time window length, This is the real-time energy consumption value. For delay The predicted energy consumption value afterward; The multi-objective weighted adaptive drift feature is measured by a drift intensity coefficient, which is calculated using the following formula: ,in The drift intensity coefficient, This is the weight vector at the current moment. For time step, To prevent tiny constants with a denominator of zero; The optimized direction phase offset index is measured by the normalized time-delay coupling amount, and the formula for calculating the normalized time-delay coupling amount is as follows: ,in This is the normalized time-delay coupling quantity. This is a time-delay coupling quantity. This is the threshold for time-delay coupling.

[0010] In a preferred embodiment, the scheduling oscillation risk index is calculated by comprehensively weighting the thermal field energy consumption lag coefficient, drift intensity coefficient, and normalized time delay coupling quantity. The calculation formula for the scheduling oscillation risk index is as follows: ,in To manage the volatility risk index These are the preset proportional coefficients for the thermal field energy consumption hysteresis coefficient, drift intensity coefficient, and normalized time-delay coupling quantity, respectively. All are greater than 0; The scheduling oscillation risk index is updated in real time during the scheduling iteration process. When the scheduling oscillation risk index continues to exceed the preset scheduling oscillation risk index threshold If the growth rate is greater than zero, then an unstable scheduling trend is determined to exist.

[0011] In a preferred embodiment, the scheduling oscillation risk index is embedded in the real-time scheduling iteration process to penalize the optimization direction, as follows: Introduce an oscillation risk penalty term into the comprehensive scheduling evaluation function: ,in This is the revised integrated scheduling evaluation function. As a risk penalty coefficient, The risk penalty function can be in quadratic form: ,in Let be the Heaviside step function, representing when Exceed The penalty is only triggered at certain times.

[0012] The technical effects and advantages of this invention are as follows: 1. This invention overcomes the bottleneck problem of "energy consumption feedback lag leading to misleading optimization" in traditional production scheduling by introducing a coupled modeling mechanism of dynamic thermal characteristics and multi-objective optimization weights into the casting production line. It achieves coordinated and self-stable scheduling of energy consumption, cycle time, and temperature control. Through the time-series acquisition of multi-source production factors and the correlation mapping of energy consumption-temperature-output, a scheduling feasibility space that can dynamically perceive changes in the thermal field is constructed, enabling the production scheduling model to have real-time response capability to unsteady thermal fields. On this basis, a multi-objective weight adaptive drift mechanism is introduced, enabling the scheduling algorithm to self-adjust weights according to the actual load and energy consumption gradient, achieving continuous self-correction of the cycle time and energy consumption optimization direction. Furthermore, by combining the optimization direction phase shift assessment and the calculation of the scheduling oscillation risk index, an active identification and suppression capability for the chain instability phenomenon of "energy consumption lag-weight drift-direction oscillation" is formed, so that the optimization process remains convergent and stable under dynamic conditions. By embedding the risk index into the scheduling iteration process, a punitive correction is implemented on the optimization direction, significantly improving the global stability, energy consumption prediction accuracy, and thermal load control consistency of the production scheduling system.

[0013] 2. This invention achieves the innovative integration of thermal field energy consumption feedback lag compensation, multi-objective weighted self-stabilizing control and nonlinear optimization phase coordination, enabling the casting production line to maintain efficient, energy-saving and stable intelligent scheduling performance even under complex thermodynamic environments. Attached Figure Description

[0014] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a flowchart of a method according to an embodiment of the present invention. Detailed Implementation

[0015] 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.

[0016] Example: Figure 1 The present invention provides an intelligent production scheduling method for a casting production line, comprising the following steps: Collect sensor data on equipment status, mold temperature, pouring cycle and energy consumption during the casting process to form a multi-source production factor sequence with timestamps; The multi-source production factor sequence is processed by feature denoising to extract parameters such as production cycle fluctuation, thermal field gradient and energy consumption distribution, and an energy consumption-temperature-output correlation mapping model is established. A dynamic feasibility map is constructed based on the association mapping model, which transforms the processing capacity, heat load and energy consumption feedback of each equipment station into node states, forming a scheduling feasibility space with time-varying weights. A multi-objective weight adaptive drift mechanism is introduced in the scheduling feasible space. By monitoring real-time production load, equipment health and energy consumption gradient, the objective function weight vector is dynamically adjusted to form a self-correcting optimization direction constraint structure. The optimization direction constraint structure is input into the nonlinear optimizer to establish the optimization direction phase offset evaluation function. By calculating the time delay coupling between the target gradient change and energy consumption feedback, the potential triggering conditions for scheduling direction oscillation are identified. By comprehensively weighting the thermal field energy consumption feedback lag information, the multi-objective weight adaptive drift characteristics, and the optimization direction phase offset index, the scheduling oscillation risk index is calculated, thereby realizing the dynamic capture of unstable scheduling trends. By embedding the scheduling oscillation risk index into the real-time scheduling iteration process, the optimization direction is penalized and corrected.

[0017] Collect sensor data on equipment status, mold temperature, pouring cycle and energy consumption during the casting process to form a multi-source production factor sequence with timestamps; In this embodiment of the invention, heterogeneous sensor channel sets are configured for the main equipment nodes in the casting production line (including melting furnaces, casting machines, sand mold cooling devices, conveying mechanisms, etc.). (n=4), where each channel corresponds to a different physical quantity: Equipment operating status signals (start / stop, current, voltage, pouring cycle time); : Mold temperature field signal (mold temperature array or infrared thermometry); : Casting time signal (time counter or visual detection); Energy consumption sensor signals (electricity meter, flow meter, or calorimeter); To avoid timing errors caused by clock drift between different signal sampling clocks, a unified master clock synchronization mechanism is adopted: ,in This refers to the clock offset at the acquisition end. The signal sampling time at the acquisition end. This refers to the signal sampling time at the acquisition end after clock synchronization; different sensor channels have different sampling frequencies, and interpolation resampling is used to unify all signals to the reference sampling frequency. This ensures that multi-source signals are computationally comparable at the same time scale; The sliding window consistency check algorithm is used to identify outlier data points: ,in This is a value used to assess data anomalies. For sampled signal data, The mean of the signal data points. The standard deviation of the signal data points; when ( When the data anomaly assessment threshold is set to a preset value, it is marked as an anomaly and interpolation based on neighborhood averaging is performed to ensure sequence continuity. Each signal, after time synchronization and interpolation correction, will be assigned a unified timestamp. Alignment is performed to generate structured multi-source sequences: And encapsulate it as a multi-source production factor vector flow; Continuous structured data streams are accumulated over time to form a multi-source production factor sequence: .

[0018] It should be noted that the above formulas are all dimensionless calculations. Commonly used methods for removing dimensions include Min-Max normalization and Z-Score standardization, which will not be elaborated here.

[0019] The multi-source production factor sequence is processed by feature denoising to extract parameters such as production cycle fluctuation, thermal field gradient and energy consumption distribution, and an energy consumption-temperature-output correlation mapping model is established. In this embodiment of the invention, sliding window mean filtering is performed on the aligned signal of each channel to denoise: ,in The sampled signal after denoising. To adjust the sliding window size, The time index within the sliding window; The extraction of production cycle fluctuations includes based on the pouring cycle signal. Extracting beat intervals : And calculate the volatility indicator: ,in The total number of samples, The average beat interval, The amplitude of the beat fluctuation is used to describe the stability of the production rhythm; The extraction of the thermal field gradient includes based on the mold temperature array Calculate the thermal gradient: ,in , These represent the rate of temperature change along the x and y directions of the mold plane, respectively. This is the thermal field gradient, used to reflect the uniformity of the thermal field distribution; The extraction of energy consumption distribution parameters includes the analysis of energy consumption signals. Perform time series statistics and sliding window analysis: , ,in This represents the average energy consumption. For energy consumption fluctuations, For the energy consumption analysis window size, This serves as the time index within the energy consumption analysis window. The establishment of the energy consumption-temperature-output correlation mapping model includes inputting the cycle time fluctuation amplitude, thermal field gradient, and energy consumption fluctuation into a multivariate regression model: ,in For output quality indicators, These are regression coefficients, obtained through training with historical data. For residual terms; Multivariate regression models can reflect the coupled impact of energy consumption and thermal field changes on production cycle time and output, providing constraints and target references for production scheduling optimization; The output quality index is compared with the preset output quality threshold. If the output quality index is greater than the output quality threshold, a dynamic feasibility map needs to be constructed based on the correlation mapping model to provide constraints and target references for production scheduling optimization. Combine cycle time fluctuation amplitude, thermal gradient, energy consumption fluctuation, and output quality indicators into a unified time series vector: .

[0020] It should be noted that the above formulas are all dimensionless calculations. Commonly used methods for removing dimensions include Min-Max normalization and Z-Score standardization, which will not be elaborated here.

[0021] A dynamic feasibility map is constructed based on the association mapping model, which transforms the processing capacity, heat load and energy consumption feedback of each equipment station into node states, forming a scheduling feasibility space with time-varying weights. In this embodiment of the invention, the processing capacity of each equipment station is determined by the cycle time fluctuation range. The heat load of each equipment station is measured by the thermal field gradient. The energy consumption feedback for each equipment station is measured by energy consumption fluctuations. measure; Define the process path dependency matrix: This matrix describes the directed dependencies of processes on the casting production line and is used to determine the connection topology of the graph structure. For any two adjacent device nodes u and v, define the energy consumption coupling weight: ,in The energy coupling weight represents the asymmetric coupling strength of energy transfer or heat exchange. For the energy consumption fluctuation of device node u For the energy consumption fluctuation of device node v Let be the thermal gradient of device node u. Let be the thermal gradient of device node v. , This is a preset proportionality coefficient between the energy consumption difference and the heat load difference, and , All are greater than 0; It should be noted that, , The settings should be tailored to the specific circumstances. For example, an expert-empowered approach could be adopted, where experts in relevant fields are invited to determine the pre-defined proportions for each indicator through professional opinion surveys and comprehensive evaluations. , The initial value can be 0.5, 0.5; Integrate all device nodes and energy consumption coupling weights into a dynamic graph structure: ,in For a set of device nodes, For the set of process-connected edges, This is a set of energy consumption coupling weights; the graph describes the characteristics of the evolution of the relationship between energy consumption and heat load among device nodes over time. Calculate the overall feasibility score for each device node: ,in The scheduling feasibility score for the device node at the current moment is given. Let u be the cycle time fluctuation amplitude of device node u. These are preset proportional coefficients for the cycle time fluctuation amplitude, energy consumption fluctuation, and thermal gradient of device node u, respectively. All are greater than 0; It should be noted that, The settings should be tailored to the specific circumstances. For example, an expert-empowered approach could be adopted, where experts in relevant fields are invited to determine the pre-defined proportions for each indicator through professional opinion surveys and comprehensive evaluations. The initial value can be 0.5, 0.25, or 0.25. A time-varying feasible space matrix is ​​constructed based on the comprehensive feasibility score and energy consumption coupling weight. : ,in It is an energy consumption suppression and adjustment factor used to balance the coupling effect between node priority and energy consumption.

[0022] It should be noted that the above formulas are all dimensionless calculations. Commonly used methods for removing dimensions include Min-Max normalization and Z-Score standardization, which will not be elaborated here.

[0023] A multi-objective weight adaptive drift mechanism is introduced in the scheduling feasible space. By monitoring real-time production load, equipment health and energy consumption gradient, the objective function weight vector is dynamically adjusted to form a self-correcting optimization direction constraint structure. In this embodiment of the invention, an initial weight vector is established based on production scheduling objectives (including minimizing processing time, minimizing energy consumption, and balancing equipment load): ,in These are the initial weights for processing time, energy consumption, and equipment load balancing objectives, respectively. The equipment load rate at the current moment is collected through the production monitoring system. Equipment health With energy consumption gradient Calculate the dynamic feedback factor : ,in This represents the historical maximum value of the equipment load rate. This represents the historical maximum value of the energy consumption gradient. This is an empirical adjustment coefficient used to balance the influence of the three types of indicators; It should be noted that the equipment load rate is the time-averaged ratio of current power to rated power; the equipment health status reflects the level of equipment performance degradation and is typically obtained through the fusion of multi-source sensor signals. Its calculation process includes: Collect key operating parameters such as vibration signals, temperature rise signals, noise amplitude, and lubrication status; Perform feature extraction (such as vibration RMS value, spectral peak value, temperature deviation rate, etc.) on each signal and compare it with the equipment's factory baseline condition; Calculate the deviation of various features, and then form a comprehensive health index based on the weighted fusion model; The health status value is usually standardized to a range of 0 to 1, where 1 represents that the device is completely healthy and 0 represents that the device is on the verge of failure. The energy consumption gradient reflects the rate of change in energy consumption per unit time and its dynamic relationship with production load and thermal environment. Its calculation logic includes: First, real-time data such as current, voltage, and temperature of the equipment are obtained at the energy consumption monitoring node, and instantaneous power is calculated; Perform time differentiation on the continuously sampled power curves to obtain the energy consumption gradient.

[0024] Based on dynamic feedback factors Update weight vector : ; Based on the updated weight vector, the direction of scheduling evaluation optimization is redefined in the scheduling feasible space: ,in The output value of the comprehensive scheduling evaluation function reflects the overall adaptability level of the scheduling scheme; Through real-time iteration The gradient changes are used to identify the drift trend of the objective function and form an adaptive optimization direction constraint structure. ,in for The gradient change.

[0025] It should be noted that the above formulas are all dimensionless calculations. Commonly used methods for removing dimensions include Min-Max normalization and Z-Score standardization, which will not be elaborated here.

[0026] The optimization direction constraint structure is input into the nonlinear optimizer to establish the optimization direction phase offset evaluation function. By calculating the time delay coupling between the target gradient change and energy consumption feedback, the potential triggering conditions for scheduling direction oscillation are identified. The specific formula for the optimized direction phase offset evaluation function in this embodiment of the invention is as follows: ,in Let be the objective function. Represent scheduling decision variables (workstation start / stop, allocation, cycle time parameters, etc.); nonlinear optimizers (such as improved BFGS or genetic algorithms) are used to search for the globally optimal direction in a multi-objective non-convex space; During the optimization iteration process, the gradient change of the objective function at adjacent time points is recorded. : ,in For time step; The gradient change sequence is transformed into a complex signal using the Hilbert Transform, which is then used to extract the phase of the target gradient change. ,in For Hilbert transform operators, The imaginary unit, The signal is a complex analytic signal consisting of the gradient change and the imaginary part of the Hilbert transform. The phase of the target gradient change; Based on the timing difference between equipment energy consumption and scheduling response, the energy consumption feedback delay phase of the energy consumption feedback signal is defined. : ,in The dominant frequency component of energy consumption variation. Energy consumption feedback delay (reflecting the time delay from dispatch command to energy consumption response); By calculating the time delay coupling between the target gradient change and energy consumption feedback. The details are as follows: ; The time-delay coupling quantity is compared with the preset time-delay coupling quantity threshold. If the time-delay coupling quantity is greater than the time-delay coupling quantity threshold, it indicates that there is a nonlinear phase misalignment between the optimization direction and the energy consumption response, that is, there is a potential risk of oscillation triggering, and it is determined that the scheduling direction oscillation has been triggered.

[0027] It should be noted that the above formulas are all dimensionless calculations. Commonly used methods for removing dimensions include Min-Max normalization and Z-Score standardization, which will not be elaborated here.

[0028] By comprehensively weighting the thermal field energy consumption feedback lag information, the multi-objective weight adaptive drift characteristics, and the optimization direction phase offset index, the scheduling oscillation risk index is calculated, thereby realizing the dynamic capture of unstable scheduling trends. In this embodiment of the invention, if scheduling direction oscillation is triggered, the scheduling oscillation risk index is calculated by comprehensively weighting the thermal field energy consumption feedback lag information, the multi-objective weight adaptive drift characteristics, and the optimized direction phase offset index. The thermal field energy consumption feedback lag information is measured by the thermal field energy consumption lag coefficient, and the calculation formula for the thermal field energy consumption lag coefficient is as follows: ,in The thermal field energy consumption lag coefficient, To calculate the time window length, This is the real-time energy consumption value. For delay The predicted energy consumption value is obtained after the operation. The thermal field energy consumption lag coefficient reflects the degree of lag in energy consumption response relative to the dispatch command. The larger the value, the stronger the thermal inertia of the system, which is more likely to cause delayed oscillations.

[0029] The multi-objective weighted adaptive drift feature is measured by a drift intensity coefficient, which is calculated using the following formula: ,in The drift intensity coefficient, This is the weight vector at the current moment. For time step, To prevent the use of small constants with denominators of zero (generally taken as...) ); It is used to characterize the dynamic trade-off fluctuations of multi-objective optimization under different constraints. The more violent the drift, the more likely the system is to generate competitive oscillations between objectives.

[0030] The optimized direction phase offset index is measured by the normalized time-delay coupling amount, and the formula for calculating the normalized time-delay coupling amount is as follows: ,in This is the normalized time-delay coupling quantity. This is a time-delay coupling quantity. The threshold for time-delay coupling; The scheduling oscillation risk index is calculated by comprehensively weighting the thermal field energy consumption lag coefficient, drift intensity coefficient, and normalized time delay coupling quantity. The formula for calculating the scheduling oscillation risk index is as follows: ,in To manage the volatility risk index These are the preset proportional coefficients for the thermal field energy consumption hysteresis coefficient, drift intensity coefficient, and normalized time-delay coupling quantity, respectively. All are greater than 0; It should be noted that, The settings should be tailored to the specific circumstances. For example, an expert-empowered approach could be adopted, where experts in relevant fields are invited to determine the pre-defined proportions for each indicator through professional opinion surveys and comprehensive evaluations. The initial value can be 0.3, 0.3, or 0.4; The scheduling oscillation risk index is updated in real time during the scheduling iteration process. When the scheduling oscillation risk index continues to exceed the preset scheduling oscillation risk index threshold And the growth rate ( If the value is greater than zero, it is determined that there is an unstable scheduling trend.

[0031] It should be noted that the above formulas are all dimensionless calculations. Commonly used methods for removing dimensions include Min-Max normalization and Z-Score standardization, which will not be elaborated here.

[0032] By embedding the scheduling oscillation risk index into the real-time scheduling iteration process, the optimization direction is penalized and corrected.

[0033] In this embodiment, an oscillation risk penalty term is introduced into the comprehensive scheduling evaluation function: ,in This is the revised integrated scheduling evaluation function. This is a risk penalty coefficient used to adjust the weighting of risk impact. The risk penalty function can be in quadratic form: ,in Let be the Heaviside step function, representing when Exceed The penalty is only triggered at certain times.

[0034] It should be noted that the above formulas are all dimensionless calculations. Commonly used methods for removing dimensions include Min-Max normalization and Z-Score standardization, which will not be elaborated here.

[0035] This invention overcomes the bottleneck problem of "energy consumption feedback lag leading to misleading optimization" in traditional production scheduling by introducing a coupled modeling mechanism of dynamic thermal field characteristics and multi-objective optimization weights into the casting production line. It achieves coordinated and self-stabilizing scheduling of energy consumption, cycle time, and temperature control. Through the time-series acquisition of multi-source production factors and the energy consumption-temperature-output correlation mapping, a scheduling feasibility space that can dynamically sense changes in the thermal field is constructed, enabling the production scheduling model to have real-time response capabilities to unsteady thermal fields. On this basis, a multi-objective weight adaptive drift mechanism is introduced, enabling the scheduling algorithm to self-adjust weights according to the actual load and energy consumption gradient, achieving continuous self-correction of the cycle time and energy consumption optimization direction. Furthermore, by combining the optimization direction phase shift assessment and the calculation of the scheduling oscillation risk index, an active identification and suppression capability is formed for the chain instability phenomenon of "energy consumption lag-weight drift-direction oscillation" is formed, so that the optimization process remains convergent and stable under dynamic conditions. By embedding the risk index into the scheduling iteration process, a punitive correction is implemented on the optimization direction, significantly improving the global stability, energy consumption prediction accuracy, and thermal load control consistency of the production scheduling system.

[0036] This invention achieves an innovative fusion of thermal field energy consumption feedback lag compensation, multi-objective weighted self-stabilizing control, and nonlinear optimization phase coordination, enabling the casting production line to maintain efficient, energy-saving, and stable intelligent scheduling performance even under complex thermodynamic environments.

[0037] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0038] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0039] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An intelligent production scheduling method for a casting processing production line, characterized in that: Includes the following steps: Collect sensor data on equipment status, mold temperature, pouring cycle and energy consumption during the casting process to form a multi-source production factor sequence with timestamps; The multi-source production factor sequence is processed by feature denoising to extract parameters such as production cycle fluctuation, thermal field gradient and energy consumption distribution, and an energy consumption-temperature-output correlation mapping model is established. A dynamic feasibility map is constructed based on the association mapping model, which transforms the processing capacity, heat load and energy consumption feedback of each equipment station into node states, forming a scheduling feasibility space with time-varying weights. A multi-objective weight adaptive drift mechanism is introduced in the scheduling feasible space. By monitoring real-time production load, equipment health and energy consumption gradient, the objective function weight vector is dynamically adjusted to form a self-correcting optimization direction constraint structure. The optimization direction constraint structure is input into the nonlinear optimizer to establish the optimization direction phase offset evaluation function. By calculating the time delay coupling between the target gradient change and energy consumption feedback, the potential triggering conditions for scheduling direction oscillation are identified. By comprehensively weighting the thermal field energy consumption feedback lag information, the multi-objective weight adaptive drift characteristics, and the optimization direction phase offset index, the scheduling oscillation risk index is calculated, thereby realizing the dynamic capture of unstable scheduling trends. By embedding the scheduling oscillation risk index into the real-time scheduling iteration process, the optimization direction is penalized and corrected.

2. The intelligent production scheduling method for a casting processing production line according to claim 1, characterized in that: The multi-source production factor sequences are processed through feature denoising to extract parameters such as production cycle fluctuation, thermal gradient, and energy consumption distribution. An energy consumption-temperature-output correlation mapping model is then established, as detailed below: Perform sliding window mean filtering for noise reduction on the aligned signal of each channel: ,in The sampled signal after denoising. To adjust the sliding window size, The time index within the sliding window; The extraction of production cycle fluctuations includes based on the pouring cycle signal. Extracting beat intervals : And calculate the volatility indicator: ,in The total number of samples, The average beat interval, For the amplitude of rhythm fluctuation The extraction of the thermal field gradient includes based on the mold temperature array Calculate the thermal gradient: ,in , These represent the rate of temperature change along the x and y directions of the mold plane, respectively. For thermal field gradient; The extraction of energy consumption distribution parameters includes the analysis of energy consumption signals. Perform time series statistics and sliding window analysis: , ,in This represents the average energy consumption. For energy consumption fluctuations, For the energy consumption analysis window size, This serves as the time index within the energy consumption analysis window. The establishment of the energy consumption-temperature-output correlation mapping model includes inputting the cycle time fluctuation amplitude, thermal field gradient, and energy consumption fluctuation into a multivariate regression model: ,in For output quality indicators, For regression coefficients, For residual terms; The output quality index is compared with the preset output quality threshold. If the output quality index is greater than the output quality threshold, a dynamic feasibility map needs to be constructed based on the correlation mapping model to provide constraints and target references for production scheduling optimization. Combine cycle time fluctuation amplitude, thermal gradient, energy consumption fluctuation, and output quality indicators into a unified time series vector: .

3. The intelligent production scheduling method for a casting processing production line according to claim 2, characterized in that: A dynamic feasibility map is constructed based on the association mapping model, which transforms the processing capacity, thermal load, and energy consumption feedback of each equipment station into node states, forming a scheduling feasibility space with time-varying weights, as detailed below: The processing capacity of each equipment station is determined by the cycle time fluctuation. The heat load of each equipment station is measured by the thermal field gradient. The energy consumption feedback for each equipment station is measured by energy consumption fluctuations. measure; Define the process path dependency matrix: ; For any two adjacent device nodes u and v, define the energy consumption coupling weight: ,in For energy consumption coupling weights, For the energy consumption fluctuation of device node u For the energy consumption fluctuation of device node v Let be the thermal gradient of device node u. Let be the thermal gradient of device node v. , This is a preset proportionality coefficient between the energy consumption difference and the heat load difference, and , All are greater than 0; Integrate all device nodes and energy consumption coupling weights into a dynamic graph structure: ,in For a set of device nodes, For the set of process-connected edges, The set of energy consumption coupling weights is used to calculate the comprehensive feasibility score for each device node. ,in The scheduling feasibility score for the device node at the current moment is given. Let u be the cycle time fluctuation amplitude of device node u. These are preset proportional coefficients for the cycle time fluctuation amplitude, energy consumption fluctuation, and thermal gradient of device node u, respectively. All are greater than 0; A time-varying feasible space matrix is ​​constructed based on the comprehensive feasibility score and energy consumption coupling weight. : ,in It is an energy consumption suppression and regulation factor.

4. The intelligent production scheduling method for a casting processing production line according to claim 3, characterized in that: A multi-objective weight adaptive drift mechanism is introduced into the scheduling feasible space. By monitoring real-time production load, equipment health, and energy consumption gradient, the objective function weight vector is dynamically adjusted to form a self-correcting optimization direction constraint structure, as follows: Based on the production scheduling objective, establish an initial weight vector: ,in These are the initial weights for processing time, energy consumption, and equipment load balancing objectives, respectively. The equipment load rate at the current moment is collected through the production monitoring system. Equipment health With energy consumption gradient Calculate the dynamic feedback factor : ,in This represents the historical maximum value of the equipment load rate. This represents the historical maximum value of the energy consumption gradient. This is an empirical adjustment coefficient; Based on dynamic feedback factors Update weight vector : ; Based on the updated weight vector, the direction of scheduling evaluation optimization is redefined in the scheduling feasible space: ,in This is the output value of the comprehensive scheduling evaluation function; Through real-time iteration The gradient changes are used to identify the drift trend of the objective function and form an adaptive optimization direction constraint structure. ,in for The gradient change.

5. The intelligent production scheduling method for a casting processing production line according to claim 4, characterized in that: The optimization direction constraint structure is input into the nonlinear optimizer to establish the optimization direction phase offset evaluation function. By calculating the time delay coupling between the target gradient change and energy consumption feedback, the potential triggering conditions for scheduling direction oscillations are identified, as follows: The specific formula for the optimized direction phase offset evaluation function is as follows: ,in Let be the objective function. Represents the scheduling decision variables; During the optimization iteration process, the gradient change of the objective function at adjacent time points is recorded. : ,in For time step; The gradient change sequence is transformed into a complex signal using the Hilbert transform, which is then used to extract the phase of the target gradient change. ,in For Hilbert transform operators, The imaginary unit, The signal is a complex analytic signal consisting of the gradient change and the imaginary part of the Hilbert transform. The phase of the target gradient change; Based on the timing difference between equipment energy consumption and scheduling response, the energy consumption feedback delay phase of the energy consumption feedback signal is defined. : ,in The dominant frequency component of energy consumption variation. Energy consumption feedback time delay; By calculating the time delay coupling between the target gradient change and energy consumption feedback. The details are as follows: ; The time-delay coupling amount is compared with the preset time-delay coupling amount threshold. If the time-delay coupling amount is greater than the time-delay coupling amount threshold, it is determined that the scheduling direction oscillation has been triggered.

6. The intelligent production scheduling method for a casting processing production line according to claim 5, characterized in that: If a scheduling oscillation is triggered, the scheduling oscillation risk index is calculated by comprehensively weighting the thermal field energy consumption feedback lag information, the multi-objective weight adaptive drift characteristics, and the optimized direction phase offset index. The thermal field energy consumption feedback lag information is measured by the thermal field energy consumption lag coefficient, and the calculation formula for the thermal field energy consumption lag coefficient is as follows: ,in The thermal field energy consumption lag coefficient, To calculate the time window length, This is the real-time energy consumption value. For delay The predicted energy consumption value afterward; The multi-objective weighted adaptive drift feature is measured by a drift intensity coefficient, which is calculated using the following formula: ,in The drift intensity coefficient, This is the weight vector at the current moment. For time step, To prevent tiny constants with a denominator of zero; The optimized direction phase offset index is measured by the normalized time-delay coupling amount, and the formula for calculating the normalized time-delay coupling amount is as follows: ,in This is the normalized time-delay coupling quantity. This is a time-delay coupling quantity. This is the threshold for time-delay coupling.

7. The intelligent production scheduling method for a casting processing production line according to claim 6, characterized in that: The scheduling oscillation risk index is calculated by comprehensively weighting the thermal field energy consumption lag coefficient, drift intensity coefficient, and normalized time delay coupling quantity. The formula for calculating the scheduling oscillation risk index is as follows: ,in To manage the volatility risk index These are the preset proportional coefficients for the thermal field energy consumption hysteresis coefficient, drift intensity coefficient, and normalized time-delay coupling quantity, respectively. All are greater than 0; The scheduling oscillation risk index is updated in real time during the scheduling iteration process. When the scheduling oscillation risk index continues to exceed the preset scheduling oscillation risk index threshold If the growth rate is greater than zero, then an unstable scheduling trend is determined to exist.

8. The intelligent production scheduling method for a casting processing production line according to claim 7, characterized in that: The scheduling oscillation risk index is embedded in the real-time scheduling iteration process to penalize the optimization direction, as follows: Introduce an oscillation risk penalty term into the comprehensive scheduling evaluation function: ,in This is the revised integrated scheduling evaluation function. As a risk penalty coefficient, The risk penalty function can be in quadratic form: ,in Let be the Heaviside step function, representing when Exceed The penalty is only triggered at certain times.

Images