Intelligent production scheduling method and system for multi-process continuous production of automobile structural parts

Abstract

This invention relates to the fields of intelligent manufacturing and industrial internet technology, and discloses an intelligent scheduling method and system for the continuous production of automotive structural components across multiple processes. The method includes: acquiring equipment status data and material flow location data to obtain a current production state vector; obtaining a preliminary production schedule using a genetic algorithm based on the current production state vector; generating adjustment demand signals based on process time fluctuation indicators; obtaining an optimized production sequence using a simulated annealing algorithm based on the adjustment demand signals; embedding emergency order response demands based on production disturbance factors to obtain a final execution plan; loading the final execution plan into a preset discrete event simulation engine to generate virtual work trajectories, comparing and generating deviation correction parameters, and parsing to generate a production instruction sequence. This method can improve the dynamic adaptability and execution accuracy of the scheduling scheme.

Description

Technical Field

[0001] This invention relates to the fields of intelligent manufacturing and industrial internet technology, and in particular to an intelligent scheduling method and system for continuous multi-process production of automotive structural components. Background Technology

[0002] Currently, the automotive manufacturing industry is at a critical juncture in its transformation towards large-scale personalized customization. Body-in-white structural components (such as longitudinal beams and crossbeams) are core parts, and their production involves the continuous connection of multiple processes, including stamping, welding, painting, and final assembly. In a mixed-line production model with multiple varieties and small batches, ensuring smooth flow between processes is a crucial factor determining the vehicle delivery cycle.

[0003] In existing technologies, production scheduling primarily relies on manual experience rules or static priority ranking methods based on fixed time windows. This approach typically assumes that process operation times are standard and fixed, pre-setting the start time and resource allocation for each process, attempting to achieve smooth production line operation under ideal conditions. However, in-depth industrial data analysis of production sites reveals that the actual production environment is fraught with uncertainty. Process operation times are not static but are influenced by a combination of real-time factors, including fluctuations in equipment operating status, differences in mold changeover time, personnel skill levels, and the timeliness of material arrival. These factors are interconnected and complex. For example, a brief delay caused by a minor equipment malfunction at a preceding welding station, without a dynamic adjustment mechanism, can quickly trigger material shortages and waiting times on downstream painting lines, or force excessive accumulation of work-in-process in upstream processes, leading to a domino effect collapse of production rhythm. Furthermore, when faced with urgent customer orders, this static scheduling method often struggles to respond quickly without disrupting the overall rhythm, frequently falling into the dilemma of sacrificing equipment utilization or creating quality risks in order to meet delivery deadlines.

[0004] In summary, existing technologies suffer from a serious disconnect between planned production schedules and actual execution, leading to wasted production resources and delivery delays. Summary of the Invention

[0005] This invention provides an intelligent scheduling method and system for the continuous production of automotive structural components across multiple processes, in order to solve the problem in the prior art where the predetermined production schedule is seriously out of sync with the actual execution status, leading to waste of production resources and delays in delivery.

[0006] Firstly, in order to solve the above-mentioned technical problems, the present invention provides an intelligent scheduling method for the continuous production of automotive structural components through multiple processes, comprising: Acquire equipment status data and material flow location data for each process in the production line, and perform preprocessing to obtain the current production status vector; Based on the current production state vector, a genetic algorithm is used to iteratively solve the multi-process tasks to obtain a preliminary production schedule. Obtain the process time fluctuation index in the preliminary production schedule. If the process time fluctuation index exceeds the preset fluctuation threshold, mark the risk process, identify the resource conflict node associated with the risk process, and determine the interference propagation path. Calculate the buffer time margin of the interference propagation path, and generate an adjustment demand signal containing the reordering priority based on the proportion of the buffer time margin consumed. The initial temperature parameters of the simulated annealing algorithm are set according to the rearrangement priority in the adjusted demand signal; based on the initial temperature parameters, a neighborhood transformation is performed on the preliminary production schedule, and an iterative search is performed using a preset multi-objective energy function to obtain an optimized production sequence; Obtain emergency order insertion response requirements, extract idle process time windows from the optimized production sequence, and obtain the current material availability status; calculate the production disturbance factor by combining the idle process time windows and the material availability status, and embed the emergency order insertion response requirements into the optimized production sequence based on the production disturbance factor to determine the final execution plan; Acquire real-time sensor data streams from the site; load the final execution plan into a preset discrete event simulation engine to generate a virtual work trajectory, and compare the virtual work trajectory with the sensor data stream to generate deviation correction parameters; parse the final execution plan to generate a production instruction sequence.

[0007] Secondly, the present invention provides an intelligent scheduling system for the continuous multi-process production of automotive structural components, comprising: The data fusion module is used to acquire equipment status data and material flow location data for each process in the production line, and to preprocess the data to obtain the current production status vector. The initial production scheduling module is used to iteratively solve multi-process tasks using a genetic algorithm based on the current production state vector to obtain a preliminary production schedule. The interference detection module is used to obtain the process time fluctuation index in the preliminary production schedule. If the process time fluctuation index exceeds the preset fluctuation threshold, the risk process is marked, the resource conflict node associated with the risk process is identified, and the interference propagation path is determined. The buffer time margin of the interference propagation path is calculated, and an adjustment demand signal containing the reordering priority is generated according to the proportion of the buffer time margin consumed. The dynamic optimization module is used to set the initial temperature parameters of the simulated annealing algorithm according to the rearrangement priority in the adjustment demand signal; perform neighborhood transformation on the preliminary production schedule based on the initial temperature parameters; and perform iterative search using a preset multi-objective energy function to obtain an optimized production sequence. The order insertion response module is used to obtain urgent order insertion response requirements, extract idle process time windows from the optimized production sequence, and obtain the current material availability status; calculate the production disturbance factor by combining the idle process time windows and the material availability status, and embed the urgent order insertion response requirements into the optimized production sequence based on the production disturbance factor to determine the final execution plan; The feedback correction module is used to acquire real-time sensor data streams collected on-site; load the final execution plan into a preset discrete event simulation engine to generate a virtual work trajectory, and compare the virtual work trajectory with the sensor data stream to generate deviation correction parameters; and parse the final execution plan to generate a production instruction sequence.

[0008] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention determines the interference propagation path by acquiring process time fluctuation indicators and identifying resource conflict nodes, and then generates differentiated adjustment demand signals based on the proportion of buffer time margin consumed. This invention breaks the limitation of traditional production scheduling that only focuses on the delay of a single process, and can accurately predict the domino-like chain effect of fluctuations on downstream processes from a global perspective; by quantifying the degree of buffer margin consumption to classify and define risks, it changes from reactive firefighting to proactive blocking, effectively preventing local minor fluctuations from evolving into cascading congestion or shutdown of the entire production line, and improving the anti-interference capability and operational robustness of multi-process continuous production systems.

[0009] (2) This invention adaptively sets the initial temperature parameters and neighborhood transformation method of the simulated annealing algorithm by analyzing and adjusting the rearrangement priority in the demand signal, and uses a multi-objective energy function to iteratively search the preliminary production schedule. This invention constructs a hierarchical response dynamic scheduling mechanism. For high-priority (high-risk) disturbances, it matches high-temperature parameters and swaps neighborhood operations, giving the algorithm a larger search space to quickly cut off the risk propagation; for low-priority disturbances, it matches medium-temperature parameters and inserts neighborhood operations, performing local fine-tuning to reduce disturbances to the overall plan. This mechanism achieves an optimal balance between the two contradictory goals of quickly eliminating resource conflicts and maintaining stable production rhythm, avoiding production line oscillations caused by over-adjustment.

[0010] (3) This invention extracts hysteresis drift features by comparing the virtual work trajectory of the final execution plan with the field sensor data stream, and generates deviation correction parameters to dynamically update the constraint weights in the scheduling rule base. This invention establishes a closed-loop self-evolutionary mechanism of planning-execution-feedback-correction, which can distinguish and identify random errors and systematic drifts in the production process (such as long-term efficiency decline caused by equipment aging); by continuously correcting the weights of the underlying scheduling rules (such as automatically increasing the buffer time weight of specific equipment), the production scheduling model can self-iterate and optimize with changes in actual working conditions, thereby improving the feasibility of production instruction sequences and the accuracy of delivery date prediction, and reducing the frequency of manual intervention. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the intelligent scheduling method for continuous multi-process production of automotive structural parts provided in the first embodiment of the present invention; Figure 2 This is a schematic diagram of the intelligent scheduling system for continuous multi-process production of automotive structural parts provided in the second embodiment of the present invention. Detailed Implementation

[0012] 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.

[0013] Reference Figure 1 The first embodiment of the present invention provides an intelligent scheduling method for continuous multi-process production of automotive structural components, including the following steps: S11: Obtain equipment status data and material flow location data for each process in the production line, and perform preprocessing to obtain the current production status vector; S12, Based on the current production state vector, use a genetic algorithm to iteratively solve the multi-process tasks to obtain a preliminary production schedule; S13, obtain the process time fluctuation index in the preliminary production schedule. If the process time fluctuation index exceeds the preset fluctuation threshold, mark the risk process, identify the resource conflict node associated with the risk process, and determine the interference propagation path. Calculate the buffer time margin of the interference propagation path, and generate an adjustment demand signal containing the reordering priority based on the proportion of the buffer time margin consumed. S14, set the initial temperature parameters of the simulated annealing algorithm according to the rearrangement priority in the adjusted demand signal; perform neighborhood transformation on the preliminary production schedule based on the initial temperature parameters, and perform iterative search using a preset multi-objective energy function to obtain an optimized production sequence; S15, obtain the emergency order insertion response requirement, extract the idle process time window from the optimized production sequence, and obtain the current material availability status; calculate the production disturbance factor by combining the idle process time window and the material availability status, and embed the emergency order insertion response requirement into the optimized production sequence based on the production disturbance factor to determine the final execution plan; S16, acquire the real-time sensor data stream collected on site; load the final execution plan into a preset discrete event simulation engine to generate a virtual work trajectory, and compare the virtual work trajectory with the sensor data stream to generate deviation correction parameters; parse the final execution plan to generate a production instruction sequence.

[0014] In step S11, equipment status data and material flow location data for each process in the production line are acquired and preprocessed to obtain the current production status vector, including: Collect equipment status data and material flow location data with millisecond-level timestamps through a sensor network; The millisecond-level timestamps are used to time-align the device status data with the material flow location data to generate multi-source input data. Feature extraction is performed on the multi-source input data to calculate the equipment runtime feature and material waiting time feature, respectively; The equipment runtime feature and the material waiting time feature are weighted and summed according to preset importance weights to generate a fused feature matrix; The fused feature matrix is ​​vectorized to obtain the current production state vector.

[0015] In one implementation, this embodiment collects equipment status data and material flow position data with millisecond-level timestamps through a sensor network. Specifically, this embodiment builds a low-level sensor network based on industrial Ethernet. For equipment status data, the system uses an edge gateway to read key register values ​​of the equipment via an industrial standard protocol, including spindle load rate, servo motor speed, operating mode, and fault alarm codes. The sampling rate of the edge gateway is set to a preset high-frequency sampling rate (e.g., 10 Hz to 50 Hz). This sampling frequency setting follows the Nyquist sampling theorem and is set to five to ten times the main frequency of the equipment's mechanical movements, aiming to ensure that the digital signal can completely reproduce the analog waveform of the equipment's instantaneous movements without aliasing distortion. For material flow position data, RFID readers or photoelectric sensors deployed at key nodes of the conveyor belt trigger event reporting when a workpiece pallet passes by, acquiring discrete event data including workpiece code, station code, and flow status. To ensure the time sequence consistency of multi-source data, the system deploys a master clock server with a precise time protocol. All edge gateways and sensor nodes act as slave clocks to synchronize with the master server, ensuring that the time synchronization error of the entire network is controlled at the millisecond level.

[0016] It should be noted that this embodiment utilizes the millisecond-level timestamps to perform time-series alignment between the equipment status data and the material flow location data, generating multi-source input data. Given that equipment data is typically a continuous, periodic time series, while material data is a discrete, bursty event series, the two are naturally misaligned on the time axis. This embodiment employs a nearest-neighbor matching strategy based on a sliding time window to perform the alignment operation. Specifically, a preset matching tolerance window is set, with the timestamp sequence of the equipment status data as the main axis. The size of this window is set to 0.5 to 1 times the sensor sampling period (e.g., 20 milliseconds) to ensure unique data matching. For each equipment status time point, the system retrieves the most recent record in the material data stream whose timestamp falls within this matching tolerance window. If a match is successful, the two are merged; if no material update data is found, a zero-order hold strategy is adopted, i.e., the material position status from the previous moment is reused, thereby generating a dense and aligned multi-source input data matrix in the time dimension.

[0017] For example, this embodiment extracts features from the multi-source input data, calculates equipment runtime features and material waiting time features respectively, and performs weighted summation according to preset importance weights to generate a fused feature matrix. For the equipment runtime feature, the system sets a statistical window based on the current production cycle. Within this window, the system parses the equipment status bits, identifies the time segments of the processing state, sums the time segments of all running states within the window, and divides by the total window duration to obtain the normalized runtime percentage. For the material waiting time feature, the system tracks the dwell time of each workpiece at the current workstation. By extracting the arrival timestamp of the workpiece entering the current buffer and subtracting the arrival timestamp from the current system time, the real-time waiting time is obtained. To eliminate differences in physical dimensions, the system performs standardization processing on the above features. The preset importance weights are objectively calculated based on historical data using the entropy weight method. Specifically, the system selects several complete production shifts from the past to construct an evaluation matrix. The rows of this matrix represent different historical time slices, and the columns represent feature indicators such as equipment runtime and material waiting time. The system calculates the information entropy of each feature index column. The smaller the numerical fluctuation of a feature across different time slices, the higher its calculated information entropy, indicating a lower contribution of that feature to distinguishing production states. The system calculates the difference coefficient based on the complementary values ​​of the information entropy and normalizes it to obtain the objective weight of each feature. This weight is then used to perform a weighted linear combination on the current feature data, thereby generating a fusion feature matrix that objectively reflects the system load state.

[0018] It is worth noting that in this embodiment, the fused feature matrix is ​​vectorized to obtain the current production state vector. Since the fused feature matrix may be a sparse matrix containing multiple devices and multiple dimensions, direct use would lead to difficulties in convergence of the subsequent genetic algorithm. Therefore, this embodiment adopts a flattening and splicing strategy. The system extracts the fused feature values ​​of each process node sequentially according to a preset process topology order (e.g., from stamping to welding to painting), and arranges and splices them into a one-dimensional real-number vector. This vector is the current production state vector, which completely encapsulates the instantaneous physical state of the entire production line in a standardized data structure, and directly serves as the chromosome encoding environment input for the genetic algorithm in subsequent steps.

[0019] In step S12, based on the current production state vector, a genetic algorithm is used to iteratively solve the multi-process tasks to obtain a preliminary production schedule, including: A resource constraint matrix is ​​constructed based on the current production status vector. The resource constraint matrix includes the available time period of equipment and the arrival time of material batches. The multi-process task set to be scheduled for production is mapped to an initial population, where each chromosome individual in the initial population represents a combination of process and equipment allocation. Based on the resource constraint matrix, the completion time index and resource utilization rate index of each chromosome individual in the initial population are calculated, and the fitness value is calculated based on the completion time index and the resource utilization rate index. Based on the fitness value, selection, crossover, and mutation operations are performed to iteratively generate the globally optimal chromosome individual; The global optimal chromosome individual is reverse-decoded, and the start and end times of each process are deduced in combination with the resource constraint matrix to generate the preliminary production schedule containing the predetermined operation time window.

[0020] In one implementation, this embodiment constructs a resource constraint matrix based on the current production state vector. This resource constraint matrix includes equipment availability time periods and material batch arrival times. Specifically, the system first performs reverse analysis on the current production state vector output in step S11, restoring the one-dimensional vector to the physical state mapping of each process node. For equipment availability time periods, the system reads the runtime characteristics of each device in the vector. If a device is currently in a processing state, its remaining processing time (based on standard working hours minus the already run time) is calculated, and the current system time plus the remaining processing time is marked as the earliest available time point for that device. If the device is in an idle or faulty state, it is marked accordingly as the current time or the estimated repair time. For material batch arrival times, the system combines the material waiting time characteristics in the vector with a preset logistics and distribution model to calculate the absolute timestamp of the materials to be processed arriving at each process buffer. The system encapsulates the above time constraint information into a two-dimensional matrix structure, namely the resource constraint matrix. Its rows represent equipment numbers, and its columns represent occupied and idle intervals on the time axis, constituting the rigid boundary conditions for subsequent production scheduling.

[0021] It should be noted that in this embodiment, the set of multi-process tasks to be scheduled is mapped to an initial population. Each chromosome individual in the initial population represents an allocation combination of processes and equipment. This embodiment uses an integer encoding scheme based on processes to construct chromosomes. Assuming there are N processes to be scheduled, a chromosome is defined as an integer sequence of length N, where the locus corresponds to the process number, and the gene value corresponds to the allocated equipment number or processing priority. To achieve a balance between solution efficiency and solution space coverage, the size of the initial population is not arbitrarily set, but is determined based on a linear function of the total number of processes to be scheduled, for example, set to 5 to 10 times the total number of processes. During the initialization phase, the system uses a strategy of random generation and heuristic rules to generate this number of chromosome individuals. Some individuals are generated through random allocation to ensure population diversity, while others are generated based on the shortest processing time priority or earliest delivery date priority rule to improve the quality of the initial population.

[0022] For example, this embodiment calculates the completion time index and resource utilization rate index for each chromosome individual in the initial population based on the resource constraint matrix, and calculates the fitness value based on the completion time index and the resource utilization rate index. For the completion time index, the system simulates the decoding process and calculates the maximum completion time corresponding to the chromosome, i.e., the difference between the last completion time and the earliest start time among all processes. For the resource utilization rate index, the system calculates the ratio of the total effective processing time of all equipment to the total online time of the equipment. Subsequently, a fitness function is constructed, which is the weighted sum of the reciprocal of the completion time index and the resource utilization rate index. To eliminate the influence of dimensions, this embodiment normalizes the above two indices before calculation. A higher fitness value indicates higher production efficiency and less resource waste in the production scheduling scheme.

[0023] In another implementation, this embodiment performs selection, crossover, and mutation operations based on the fitness value to iteratively generate the globally optimal chromosome individual. In the selection operation, a tournament selection method is used, where a preset number of individuals are randomly selected each time for fitness comparison, retaining the winners to enter the next generation, while also retaining the elite individuals with the highest fitness in each generation for direct inheritance. In the crossover operation, a partially matched crossover operator is used to exchange gene segments of the parent chromosomes while ensuring that the process constraints do not conflict, in order to inherit superior substructures. In the mutation operation, the mutation probability is set within the range of 0.005 to 0.05 and dynamically adjusted according to the current fitness variance of the population; this range is selected to prevent high-frequency mutations from destroying the generated superior gene patterns, while maintaining population diversity to prevent premature convergence. The above process is iterated until a preset maximum number of iterations is reached or a fitness convergence threshold is met. The fitness convergence threshold is defined as the rate of change of the maximum fitness of the population being less than a preset minimum value (e.g., 1e-4) for several consecutive generations (e.g., 10 generations), at which point the individual with the highest fitness in the population is marked as the globally optimal chromosome individual.

[0024] It is worth noting that this embodiment performs reverse decoding on the globally optimal chromosome individual, and combines the resource constraint matrix to deduce the start and end times of each process, generating the preliminary production schedule containing the predetermined operation time window. The decoding process adopts an active scheduling decoding algorithm. The system scans the idle time windows of the corresponding equipment in the resource constraint matrix according to the process priority determined by the chromosome gene sequence. Under the premise of satisfying the preceding constraints of the process (i.e., the previous process must be completed) and the material arrival constraints, the current process is inserted into the earliest available idle time period of the equipment, thereby determining the planned start and end times of the process. The interval formed by these two time points is the predetermined operation time window. After traversing all processes, the final generated Gantt chart data structure containing the full-process time window plan is the preliminary production schedule.

[0025] In step S13, the process time fluctuation index in the preliminary production schedule is obtained. If the process time fluctuation index exceeds a preset fluctuation threshold, a risky process is marked, the resource conflict node associated with the risky process is identified, and the interference propagation path is determined, including: Extract the predetermined operation time window from the preliminary production schedule and obtain the historical operation data of the corresponding equipment for the process; Calculate the standard deviation or range of the historical operating data of the equipment relative to the predetermined operation time window, and determine it as the process time fluctuation index; If the process time fluctuation index is greater than the preset fluctuation threshold, the corresponding process will be marked as the risk process. Identify the equipment nodes and material nodes occupied by the risky process, determine the subsequent affected process sequence, and generate the interference propagation path.

[0026] In one implementation, this embodiment extracts the predetermined operation time window from the preliminary production schedule and obtains the historical operating data of the corresponding process equipment. Specifically, the system parses the Gantt chart data structure generated in step S12 and reads the planned duration of each process node. Subsequently, the system uses the process type and equipment number as a joint index to retrieve the actual execution records of all similar processes within a preset past period from the historical database of the manufacturing execution system. To ensure the statistical significance of the data, the system automatically filters out extreme outliers caused by equipment failure or downtime maintenance, retaining only the operation duration data under normal production conditions to form a historical sample set. Based on this, this embodiment calculates the standard deviation or range of the historical equipment operating data relative to the predetermined operation time window and determines it as the process time fluctuation index. This index quantifies the possible time deviation of the process during future execution.

[0027] It should be noted that in this embodiment, if the process time fluctuation index exceeds the preset fluctuation threshold, the corresponding process is marked as the risk process. The preset fluctuation threshold is determined by reverse derivation based on the process capability index after time parameter correction. Specifically, the system establishes a regression analysis model with the standard deviation of process operation time as the independent variable and the cumulative downtime caused by material shortages in downstream processes as the dependent variable. By fitting historical data, the system analyzes the marginal growth rate of the dependent variable as a function of the independent variable, identifying the inflection point where downtime increases sharply. The standard deviation value corresponding to this inflection point is the preset fluctuation threshold. The safety factor corresponding to this threshold is typically set between 0.6 and 0.7 (e.g., 0.618), its physical meaning being to establish an optimal empirical balance between the economic cost of buffer resource occupation and the risk of production line downtime.

[0028] For example, this embodiment identifies the equipment nodes occupied by the risky process and the material nodes they depend on, determines the subsequent affected process sequence, and generates the interference propagation path. This step uses a depth-first search strategy to construct the impact chain. The system first retrieves the immediate successor processes that need to be executed on the same equipment after the risky process. If the risky process is delayed, the equipment will be occupied for a long time, and such nodes are identified as resource conflict nodes at the equipment level. Second, the system retrieves all downstream processes that require the output of the risky process as raw materials. If the risky process is delayed, the downstream processes will stop due to material shortages, and such nodes are identified as dependent nodes at the material level. The system uses the risky process as the root node and recursively traverses downstream along the above equipment occupation relationship and material flow relationship, connecting all affected process nodes in the order of their impact time to form a directed acyclic graph path, which is the interference propagation path.

[0029] It is worth noting that in step S13, the buffer time margin of the interference propagation path is calculated, and an adjustment demand signal containing the reordering priority is generated based on the consumption ratio of the buffer time margin. This embodiment first obtains the reserved buffer time for each process on the interference propagation path, i.e., the sum of the planned process interval time and the time converted from the physical buffer capacity, and calculates the total buffer time margin. Next, based on the process time fluctuation index of the risky process, the expected delay time is estimated using the three-standard-deviation principle of the normal distribution, and the proportion of this expected delay time to the total buffer time margin is calculated to obtain the consumption ratio. If the consumption ratio exceeds a preset high-intensity threshold, it means that the current interference is very likely to breach all buffer defenses, and the system generates a first-level response signal, identifying the processes on the interference propagation path as first-priority reordering targets; if the consumption ratio is within a preset medium-intensity range, a second-level response signal is generated, identifying the relevant processes as second-priority reordering targets. Finally, the system encapsulates the information containing the above priority markers into the adjustment demand signal.

[0030] It is important to note that the preset high-intensity threshold (e.g., 0.85) is determined based on phase transition critical point analysis. The system analyzes historical production logs to construct a curve showing the relationship between buffer consumption ratio and the probability of cascading failures. By calculating the second derivative of this curve, the inflection point where the failure probability accelerates is identified, and the ratio corresponding to this inflection point is determined as the high-intensity threshold. This setting ensures that the system only triggers high-priority global reordering when the risk is about to spiral out of control, avoiding excessive resource consumption.

[0031] In step S13, the buffer time margin of the interference propagation path is calculated, and an adjustment demand signal containing the rearrangement priority is generated based on the proportion of the buffer time margin consumed, including: Obtain the reserved buffer time for each process along the interference propagation path, and calculate the total buffer time margin; The estimated delay time is calculated based on the process time fluctuation index of the risk process, and the proportion of the estimated delay time to the total buffer time is calculated to obtain the consumed proportion. If the consumption ratio exceeds the preset high-intensity threshold, a first-level response signal is generated, and the process on the interference propagation path is identified as the first priority reordering object. If the consumption ratio is within the preset medium intensity range, a secondary response signal is generated, and the process on the interference propagation path is identified as the second priority rearrangement object. The priority marker containing the first-level response signal or the second-level response signal is encapsulated as the adjustment requirement signal.

[0032] In one implementation, this embodiment obtains the reserved buffer time for each process along the interference propagation path and calculates the total buffer time margin. Specifically, the system traverses each process node in the interference propagation path, reads the time difference between its planned end time in the preliminary production schedule and the planned start time of the next immediately following process, and defines it as the planned buffer time. Simultaneously, the system combines the physical capacity of the on-site material buffer area with the current workstation cycle time to convert the physical buffer capacity into physical buffer time. The system accumulates the planned buffer time and physical buffer time for all process nodes along the path to obtain the total buffer time margin. This physical quantity represents the maximum interference energy that the production path can absorb without triggering a full-line shutdown.

[0033] In another implementation, this embodiment calculates the expected delay time based on the process time fluctuation index of the risk process, and calculates the proportion of the expected delay time to the total buffer time margin to obtain the consumed proportion.

[0034] To ensure the safety of the risk assessment, this embodiment employs a worst-case estimation method for estimating the expected delay time. Specifically, it utilizes process time fluctuation indicators (e.g., standard deviation) calculated in previous steps. ), combined with normal distribution In principle, the expected delay duration is defined as three times the volatility index (i.e. This means that the delay covers 99.73% of the potential probability of occurrence. It should be noted that using 3 times the standard deviation as the expected delay duration is a conservative estimation method based on historical statistics, which aims to cover the vast majority of potential delay scenarios and provide sufficient safety margin for risk warning and buffer calculation, rather than a precise point prediction.

[0035] The system then divides the estimated delay time by the total buffer time margin to obtain a dimensionless value between 0 and 1, which is the consumed proportion. This proportion intuitively reflects the degree to which the current interference erodes the system's resilience.

[0036] It should be noted that in this embodiment, if the consumption ratio exceeds the preset high-intensity threshold, a first-level response signal is generated, and the process on the interference propagation path is identified as the first priority reordering object.

[0037] The preset high-intensity threshold (e.g., 0.85) is not arbitrarily set, but determined based on phase transition critical point analysis. By analyzing historical production logs and statistically analyzing the correlation between buffer consumption ratio and downstream outage events, the system found that when the consumption ratio exceeds a certain critical value, the probability of cascading shutdowns increases exponentially. This critical value is then set as the high-intensity threshold.

[0038] For example, if the consumption ratio is in a preset medium intensity range, a secondary response signal is generated, and the process on the interference propagation path is identified as the second priority rearrangement object.

[0039] The preset medium-intensity range (e.g., [0.4, 0.85]) corresponds to the sub-healthy state of the system. At this time, although the interference has not completely exhausted the buffer, it has severely compressed the elasticity space of subsequent processes, which can easily cause production rhythm disorder (such as sudden fast and slow). Therefore, the secondary response signal generated by the system corresponds to a mild correction strategy, marking the relevant processes as the second highest priority, and instructing the subsequent algorithm to only perform local sequential fine-tuning (such as insertion operation) to smooth the production rhythm.

[0040] It is worth noting that in this embodiment, the priority marker containing the first-level response signal or the second-level response signal is encapsulated as the adjustment request signal. The system constructs a standardized message frame structure, which includes an event ID, a triggering source process ID, a list of affected paths, and core reordering priority code. This adjustment request signal is pushed to the subsequent scheduling optimization engine as a direct basis for setting the temperature parameters of the simulated annealing algorithm, thereby realizing a logical closed loop from risk perception to differentiated decision-making.

[0041] In step S14, the initial temperature parameters of the simulated annealing algorithm are set according to the rearrangement priority in the adjusted demand signal; based on the initial temperature parameters, a neighborhood transformation is performed on the preliminary production schedule, and an iterative search is conducted using a preset multi-objective energy function to obtain an optimized production sequence, including: If the rearrangement priority is the first priority, then the initial temperature parameter is set to a preset high temperature value, and the neighborhood exchange operation is determined to be the corresponding neighborhood transformation method. If the rearrangement priority is the second priority, then the initial temperature parameter is set to a preset medium temperature value, and the insertion neighborhood operation is determined to be the corresponding neighborhood transformation method. Construct the multi-objective energy function, which includes a device idle time term, a cumulative process delay term, and a production rhythm deviation term; Candidate sequences are generated based on the initial temperature parameters and the neighborhood transformation method. The system energy value of the candidate sequences is calculated, and the optimal solution is updated according to the preset acceptance probability criterion until the preset iteration termination condition is met, and the optimized production sequence is output.

[0042] In one implementation, if the rearrangement priority is the first priority, the initial temperature parameter is set to a preset high temperature value, and the neighborhood exchange operation is determined as the corresponding neighborhood transformation method. It should be noted that the first priority corresponds to a high-intensity interference scenario, where the original production sequence structure has been severely damaged, and it is necessary to escape the current local optimum trap. The preset high temperature value is determined by reverse derivation based on the initial acceptance probability criterion. Specifically, the system first generates several sets of neighborhood solutions through random sampling and calculates their average energy degradation value; simultaneously, it sets an initial acceptance probability target for inferior solutions (e.g., 80% to 90%), meaning that the algorithm should initially accept temporary solutions worse than the current solution with a very high probability; the system calculates the preset high temperature value by dividing the negative of the average energy degradation value by the natural logarithm of the initial acceptance probability target. The accompanying neighborhood exchange operation specifically refers to randomly selecting two non-adjacent process nodes in the current process sequence and directly exchanging their positions on the time axis. This significant structural perturbation can quickly change the topology of the sequence, assisting the algorithm in making long-distance jumps in the solution space.

[0043] In another implementation, if the rearrangement priority is the second priority, the initial temperature parameter is set to a preset mid-temperature value, and the neighbor insertion operation is determined to be the corresponding neighborhood transformation method. It should be noted that the second priority corresponds to a medium-intensity interference scenario, where the overall skeleton of the production sequence remains robust, requiring only local repair. The preset mid-temperature value is set to a value corresponding to a medium acceptance probability (e.g., around 30%), aiming to retain a certain level of exploration capability while preventing the search process from becoming too divergent. The accompanying neighbor insertion operation specifically refers to randomly removing a process node from the current process sequence and re-inserting it into another empty position in the sequence. This operation preserves the relative order of most subsequences and is mainly used to fine-tune the compactness between processes, achieving fine-tuning and convergence of existing solutions.

[0044] For example, this embodiment constructs the multi-objective energy function, which includes a device idle time term, a cumulative process delay term, and a production rhythm deviation term. This function aims to compress multi-dimensional scheduling objectives into a single scalar value; a lower energy value indicates a better solution. Specifically, the device idle time term quantifies the compactness of resource utilization, the cumulative process delay term quantifies the timeliness of delivery, and the production rhythm deviation term quantifies the stability of the production flow by statistically analyzing the variance of the time interval between adjacent workpieces. It should be noted that, given the inconsistent physical dimensions of the three indicators (e.g., idle time and delay time are time-based, while the variance of the rhythm deviation term is time-squared), direct weighting would lead to calculation distortion. Therefore, before weighted summation, this embodiment first performs dimensionless normalization (e.g., max-min standardization) on the three indicators, mapping them to a dimensionless range between zero and one. Then, using preset weighting coefficients, the normalized indicators are weighted and summed to obtain the final system energy value. The weight coefficients are calculated in real time using the entropy weight method. Specifically, an evaluation matrix is ​​constructed by selecting candidate solution sets generated from several past iterations, and the information entropy of each index column is calculated. If the degree of variation of an index is greater, it is given a higher weight, thereby objectively reflecting the importance of each index in the current solution space.

[0045] It is worth noting that this embodiment generates candidate sequences based on the initial temperature parameter and the neighborhood transformation method, calculates the system energy value of the candidate sequences, and updates the optimal solution according to a preset acceptance probability criterion until a preset iteration termination condition is met, outputting the optimized production sequence. Specifically, the iteration logic is as follows: in each iteration cycle, the algorithm generates a new candidate sequence based on the current temperature and the selected neighborhood method. If the energy value of the candidate sequence is lower than the energy value of the current sequence, the candidate sequence is directly accepted as the new solution; if the energy value of the candidate sequence is higher than the current sequence, the energy difference is calculated, and the acceptance probability is calculated based on the Metropolis criterion. This probability is obtained by calculating the natural constant raised to the power of its exponent, where the exponent is the negative of the energy difference divided by the current temperature parameter. The system generates a random number between zero and one. If this random number is less than the calculated acceptance probability, the degenerate solution is tolerated and adopted as the new solution; otherwise, the original solution remains unchanged. As iterations proceed, the temperature parameter gradually decreases according to a preset cooling rate, which (i.e., the temperature decay coefficient) is typically set between 0.90 and 0.99 to ensure that the algorithm has sufficient time to reach thermal equilibrium at each temperature level, eventually converging to the global optimum or near-optimal state. When the temperature falls below a preset termination threshold or the energy value no longer decreases significantly for several consecutive generations, the algorithm terminates and outputs the optimized production sequence.

[0046] In step S15, the production disturbance factor is calculated by combining the idle process time window and the material availability status, and the emergency order insertion response demand is embedded into the optimized production sequence based on the production disturbance factor to determine the final execution plan, including: Identify the time periods in the optimized production sequence where no tasks are assigned, and determine them as the idle process time windows; Obtain the list of materials required to respond to the emergency order request, and check the current inventory status of the list of materials to obtain the material availability status; The delay time of subsequent processes caused by embedding the emergency order response demand into each idle process time window is calculated and determined as the production disturbance factor. The time window with the smallest production disturbance factor is selected as the minimum disturbance insertion point; The emergency order response requirement is embedded into the minimum disturbance insertion point, and subsequent processes are sequentially reorganized to generate the final execution plan.

[0047] In one implementation, this embodiment identifies unassigned time periods in the optimized production sequence and determines them as the idle process time window. Specifically, the system iterates through the optimized production sequence output in step S14 and reads the planned end time and planned start time of two adjacent processes on the same equipment. The system calculates the difference between the end time of the previous process and the start time of the next process. If the difference is greater than zero, the time period is marked as the physical idle interval of the equipment. The system further filters out the rigidly occupied time reserved for equipment maintenance, shift changes, or necessary mold cooling, defines the remaining available idle interval as the idle process time window, and records its start point, end point, and corresponding equipment number.

[0048] In another implementation, this embodiment obtains the bill of materials required to respond to the emergency order and checks the current inventory status of the bill of materials to obtain the material availability status. The system first parses the bill of materials information for the emergency order response and queries the real-time inventory level and transit status of relevant materials through the interface of the enterprise resource planning or warehouse management system. If all materials are in stock, the current system time is marked as the availability time point; if there is a shortage, the estimated arrival time of the purchase order is read, and the latest estimated arrival time of all materials is marked as the availability time point. This embodiment encapsulates the availability time point as the material availability status, which determines the earliest physical time limit at which the emergency order can begin processing.

[0049] For example, this embodiment calculates the delay time of subsequent processes caused by embedding the emergency order insertion response demand into each of the idle process time windows, and determines it as the production disturbance factor. This step uses a virtual trial insertion strategy for calculation. For each candidate idle process time window, the system first compares the start point of the time window with the material matching time point, and takes the later of the two as the earliest feasible start time of the emergency order insertion. Then, the operation time of the emergency order insertion is superimposed on the earliest feasible start time to calculate the expected end time. It should be noted that, in order to quantify the impact on the whole, the system performs cascading extrapolation calculations. Specifically, for affected process nodes, the system first reads their free float time in the original optimized production sequence, which is the maximum amount of time that the process can delay starting without affecting the earliest completion time of the entire project. If the direct squeeze time is less than or equal to the free float time of the process, the impact is absorbed in the process and the transmission ends. If the direct squeeze time is greater than the free float time, the excess part is the effective extension time of the process. This effective extension time will be used as the new squeeze time and will continue to be transmitted to every subsequent process of the process, and the above judgment process will be repeated until the impact on all paths is absorbed or transmitted to the final process. The production disturbance factor is the sum of all effective extension times triggered during the simulation process.

[0050] It is worth noting that in this embodiment, the time window with the smallest production disturbance factor is selected as the minimum disturbance insertion point, and the emergency order response requirement is embedded into the minimum disturbance insertion point. Subsequent processes are then sequentially reorganized to generate the final execution plan. The system sorts the production disturbance factors calculated for all candidate time windows and selects the time window with the smallest value, i.e., the one that causes the least disruption to the original plan, as the optimal insertion position. After confirming the insertion, the system performs a chain update operation, writing the emergency order task into that position and updating the planned start and end times of all subsequent affected processes. This maintains the original logical constraints between processes, such as processing order and minimum interval, thus generating the final execution plan that includes the emergency task while maintaining the original rhythm to the greatest extent possible.

[0051] In step S16, the final execution plan is loaded into a preset discrete event simulation engine to generate a virtual job trajectory, and the virtual job trajectory is compared with the sensor data stream to generate deviation correction parameters, including: Run the discrete event simulation engine and output the virtual job trajectory containing the simulated completion time; Collect actual processing progress data on site, calculate the difference between the simulated completion time and the actual completion time, and generate an expected deviation dataset; Analyzing the expected deviation dataset, if the deviation shows a continuous cumulative trend, it is extracted as a systematic drift feature; if the deviation shows an occasional distribution, it is extracted as a random drift feature; the systematic drift feature and the random drift feature are collectively referred to as hysteresis drift features. For the systematic drift characteristics, a buffer time weight coefficient for the corresponding process is generated; for the random drift characteristics, a priority correction coefficient for the corresponding process is generated; the buffer time weight coefficient and the priority correction coefficient are combined to form the deviation correction parameter.

[0052] In one implementation, this embodiment runs the discrete event simulation engine and outputs the virtual work trajectory containing the simulated completion time. Specifically, this embodiment uses a simulation kernel built based on a next event time advancement mechanism. This kernel includes a global simulation clock, an event list, and a set of system state variables. The system parses the final execution plan determined in step S15 into a series of discrete events to be triggered, such as entry, processing start, processing end, and exit. The simulation clock does not flow uniformly but advances in a jump-like manner according to the logical order of event occurrence. Each time an event is triggered, the engine updates the state of the virtual entity according to the preset equipment process model. It should be noted that the equipment process model is not a simple set of static parameters, but a stochastic process model derived from the statistical analysis of fault records in the equipment lifecycle management system or equipment maintenance logs over the past year. This model pre-stores key operating parameters such as the standard processing cycle time, mean time between failures (MTBF), and mean time to repair for each piece of equipment. During the simulation, the system sets the MTBF to follow an exponential distribution and the repair time to follow a log-normal distribution, thereby simulating the random state changes and performance fluctuations that may occur during the future execution of the equipment. After traversing all planned tasks, the system generates a time-series log containing the expected start and end times of each process node, which is the virtual job trajectory.

[0053] It is worth noting that the parameters of the stochastic process model are obtained by analyzing the equipment's historical maintenance logs and processing records. For example, the mean time between failures (MTBF) is obtained by calculating the average of the total normal operating time of the equipment between two consecutive fault alarms; the mean time to repair (MTBT) is obtained by calculating the average of the time taken from the occurrence to the resolution of each fault. During simulation, when the virtual clock advances to the point where a certain piece of equipment begins processing, the system first randomly generates a potential fault time point for this processing based on an exponential distribution. If this time point is earlier than the planned completion time of the process, a fault event is inserted at this point, processing is paused, and the repair time is randomly generated based on a log-normal distribution. If the potential fault time point is later than the planned completion time, no fault occurs during this processing. In this way, deterministic processes are coupled with stochastic equipment events.

[0054] It should be noted that this embodiment collects actual processing progress data from the site, calculates the difference between the simulated completion time and the actual completion time, and generates an expected deviation dataset. The system uses the same sensor network interface as in the previous steps to capture the actual completion signal of the corresponding process in the physical production line in real time. The system subtracts the simulated completion time in the virtual work trajectory from the actual completion time of the same process task to obtain a signed time deviation value. To capture long-term evolution patterns, the system maintains a sliding time window of a preset length, arranging all deviation values ​​within this window in chronological order to form the expected deviation dataset for analysis. The length of this sliding time window is not arbitrarily selected, but is set to cover at least one complete production cycle or the minimum wear cycle of the equipment tool (e.g., fifty completed batches) to ensure that the sample size meets the law of large numbers requirement in statistics, thereby avoiding statistical bias caused by insufficient samples.

[0055] In another implementation, this embodiment analyzes the expected deviation dataset. If the deviation shows a continuous cumulative trend, it is extracted as a systematic drift feature; if the deviation shows an occasional distribution, it is extracted as a random drift feature. This step aims to mathematically decouple the causes of the deviation. For systematic drift features, this embodiment uses linear regression analysis to fit a straight line to the deviation data within the sliding window. If the absolute value of the slope of the fitted line is greater than a preset trend threshold, and the goodness of fit meets statistical requirements, it is determined that the production line has structural efficiency degradation caused by tool wear or equipment aging, and the slope value of the fitted line is extracted as the systematic drift feature. The trend threshold is set to 0.01, which represents a significance level of 1% efficiency degradation per 100 processing cycles. For random drift features, this embodiment uses statistical dispersion analysis. If the mean of the deviation data is close to zero, but the standard deviation exceeds a preset dispersion threshold, it is determined that the production line is affected by uncontrollable environmental disturbances such as fluctuations in manual clamping speed, and the standard deviation value is extracted as the random drift feature. The dispersion threshold is set to three times the standard deviation under normal historical processing conditions for this process, i.e., exceeding the 99.73% confidence interval. The systematic drift feature and the random drift feature are collectively referred to as hysteresis drift features in subsequent processing.

[0056] For example, this embodiment generates deviation correction parameters based on the hysteresis drift characteristics and updates the constraint weights in the preset scheduling rule base using the deviation correction parameters. This embodiment adopts a strategy of parallel execution and evolution. In the evolution thread, for the identified systematic drift, the system calculates the buffer time weight coefficient using the linear mapping formula, that is, the new weight is equal to the old weight multiplied by the correction factor, which is composed of a product of a gain coefficient and the drift slope, thereby automatically extending the reserved window of the relevant process in future production scheduling; for random drift, the system calculates the priority correction coefficient using the differential feedback formula, that is, the new weight is equal to the old weight plus the product of the proportional coefficient and the portion of the current standard deviation exceeding the target value, thereby improving the material preemption right of the relevant process in future resource allocation. These corrected weights are written into the scheduling rule base and are explicitly marked as the benchmark parameters for calculating the weighted sum of features or generating the initial population in the next production scheduling cycle, thereby realizing the adaptive iteration of the production scheduling model, making it increasingly closer to the real physical working conditions as the production cycle progresses.

[0057] It is worth noting that in this embodiment, the final execution plan determined in step S15 is directly parsed in the parallel thread updating the weights to generate the production instruction sequence. To ensure the continuity and stability of this production, the system directly locks the final execution plan generated in step S15 and converts it into an instruction format recognizable by the underlying equipment. For example, G-code program call instructions are generated for CNC machine tools, material handling task instructions are generated for AGV carts, and electronic work instructions are generated for manual workstations. These instructions are encapsulated into a queue in chronological order, which is the production instruction sequence, and is sent to the production site for execution via a message queue telemetry transmission protocol.

[0058] For example, this embodiment also establishes a weight stability verification mechanism. Although the current production directly executes the locked plan, the system will perform virtual simulations in the background using the updated weights. The system calculates the deviation between the hypothetical plan generated based on the new weights and the current actual execution plan in terms of the expected completion time of key processes, and generates an evolutionary stability index. If the index remains within a preset convergence range for several consecutive periods, it indicates that the system's scheduling rule base has reached a steady state that matches the current physical production line status through adaptive adjustment. At this time, the system can choose to solidify the current weights to reduce subsequent computational overhead. If the index fluctuates drastically, it prompts the system to continue to maintain a highly sensitive weight update strategy in the next period.

[0059] In summary, this invention constructs a high-precision initial production scheduling scheme through multi-source data fusion and genetic algorithms, and utilizes a risk propagation model based on buffer time consumption ratio to achieve graded early warning of production disturbances. Through a priority-driven adaptive simulated annealing algorithm and a minimum disturbance insertion strategy, it achieves rapid response and flexible repair to sudden anomalies and urgent demands in a dynamic environment. Furthermore, by leveraging discrete event simulation and residual analysis of sensor data, a closed-loop feedback mechanism of parallel execution and evolution is established, transforming hysteresis drift characteristics into scheduling rule optimization parameters for the next cycle. This achieves intelligent control throughout the entire process, from static planning to dynamic adaptation and then to autonomous evolution, effectively solving the industry problem of plans failing to keep up with changes in automotive structural component production, and improving the operational robustness and on-time delivery rate of the production line.

[0060] It is worth noting that this invention is based on a digital production management platform for stamping dies. This platform is a comprehensive system integrating multiple technologies, covering the entire process of digital production management from design to manufacturing to inspection, forming a closed-loop system from process simulation, production scheduling, machining control to quality inspection. The design and planning stage includes a die surface correction system, which corrects the die surface using simulation and measured data to guide the final die surface finishing. The manufacturing stage includes an intelligent production scheduling and dispatching system, which improves the dynamic adaptability and execution accuracy of the scheduling plan through dynamic scheduling, order insertion response, and production scheduling optimization; and an adaptive machining system, which improves the accuracy and efficiency of die remachining through machine measurement, error compensation, and adaptive machining. The inspection stage includes a die cutting edge wear life prediction system, which achieves high-precision cutting edge wear life prediction. The prediction results affect the equipment health in the production scheduling system, facilitating preventative maintenance scheduling; and a drawing die critical surface life prediction system, which predicts the remaining life by monitoring the health status of the die body, providing support for maintenance plans.

[0061] Reference Figure 2 The second embodiment of the present invention provides an intelligent scheduling system for the continuous production of automotive structural components across multiple processes, comprising: The data fusion module is used to acquire equipment status data and material flow location data for each process in the production line, and to preprocess the data to obtain the current production status vector. The initial production scheduling module is used to iteratively solve multi-process tasks using a genetic algorithm based on the current production state vector to obtain a preliminary production schedule. The interference detection module is used to obtain the process time fluctuation index in the preliminary production schedule. If the process time fluctuation index exceeds the preset fluctuation threshold, the risk process is marked, the resource conflict node associated with the risk process is identified, and the interference propagation path is determined. The buffer time margin of the interference propagation path is calculated, and an adjustment demand signal containing the reordering priority is generated according to the proportion of the buffer time margin consumed. The dynamic optimization module is used to set the initial temperature parameters of the simulated annealing algorithm according to the rearrangement priority in the adjustment demand signal; perform neighborhood transformation on the preliminary production schedule based on the initial temperature parameters; and perform iterative search using a preset multi-objective energy function to obtain an optimized production sequence. The order insertion response module is used to obtain urgent order insertion response requirements, extract idle process time windows from the optimized production sequence, and obtain the current material availability status; calculate the production disturbance factor by combining the idle process time windows and the material availability status, and embed the urgent order insertion response requirements into the optimized production sequence based on the production disturbance factor to determine the final execution plan; The feedback correction module is used to acquire real-time sensor data streams collected on-site; load the final execution plan into a preset discrete event simulation engine to generate a virtual work trajectory, and compare the virtual work trajectory with the sensor data stream to generate deviation correction parameters; and parse the final execution plan to generate a production instruction sequence.

[0062] It should be noted that the intelligent scheduling system for continuous multi-process production of automotive structural parts provided in this embodiment of the invention is used to execute all process steps of the intelligent scheduling method for continuous multi-process production of automotive structural parts in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0063] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0064] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. An intelligent scheduling method for continuous multi-process production of automotive structural components, characterized in that, include: Acquire equipment status data and material flow location data for each process in the production line, and perform preprocessing to obtain the current production status vector; Based on the current production state vector, a genetic algorithm is used to iteratively solve the multi-process tasks to obtain a preliminary production schedule. Obtain the process time fluctuation index in the preliminary production schedule. If the process time fluctuation index exceeds the preset fluctuation threshold, mark the risk process, identify the resource conflict node associated with the risk process, and determine the interference propagation path. Calculate the buffer time margin of the interference propagation path, and generate an adjustment demand signal containing the reordering priority based on the proportion of the buffer time margin consumed. The initial temperature parameters of the simulated annealing algorithm are set according to the rearrangement priority in the adjusted demand signal; based on the initial temperature parameters, a neighborhood transformation is performed on the preliminary production schedule, and an iterative search is performed using a preset multi-objective energy function to obtain an optimized production sequence; Obtain emergency order insertion response requirements, extract idle process time windows from the optimized production sequence, and obtain the current material availability status; calculate the production disturbance factor by combining the idle process time windows and the material availability status, and embed the emergency order insertion response requirements into the optimized production sequence based on the production disturbance factor to determine the final execution plan; Acquire real-time sensor data streams from the field; The final execution plan is loaded into a preset discrete event simulation engine to generate a virtual job trajectory, and the virtual job trajectory is compared with the sensor data stream to generate deviation correction parameters; the final execution plan is parsed to generate a production instruction sequence.

2. The intelligent scheduling method for continuous multi-process production of automotive structural components according to claim 1, characterized in that, The process of acquiring equipment status data and material flow location data for each process in the production line, and preprocessing them to obtain the current production status vector, includes: Collect equipment status data and material flow location data with millisecond-level timestamps through a sensor network; The millisecond-level timestamps are used to time-align the device status data with the material flow location data to generate multi-source input data. Feature extraction is performed on the multi-source input data to calculate the equipment runtime feature and material waiting time feature, respectively; The equipment runtime feature and the material waiting time feature are weighted and summed according to preset importance weights to generate a fused feature matrix; The fused feature matrix is ​​vectorized to obtain the current production state vector.

3. The intelligent scheduling method for continuous multi-process production of automotive structural components according to claim 2, characterized in that, The step of using a genetic algorithm to iteratively solve multi-process tasks based on the current production state vector to obtain a preliminary production schedule includes: A resource constraint matrix is ​​constructed based on the current production status vector. The resource constraint matrix includes the available time period of equipment and the arrival time of material batches. The multi-process task set to be scheduled for production is mapped to an initial population, where each chromosome individual in the initial population represents a combination of process and equipment allocation. Based on the resource constraint matrix, the completion time index and resource utilization rate index of each chromosome individual in the initial population are calculated, and the fitness value is calculated based on the completion time index and the resource utilization rate index. Based on the fitness value, selection, crossover, and mutation operations are performed to iteratively generate the globally optimal chromosome individual; The global optimal chromosome individual is reverse-decoded, and the start and end times of each process are deduced in combination with the resource constraint matrix to generate the preliminary production schedule containing the predetermined operation time window.

4. The intelligent scheduling method for continuous multi-process production of automotive structural components according to claim 3, characterized in that, The process of obtaining the process time fluctuation index in the preliminary production schedule, and marking the risky process if the process time fluctuation index exceeds a preset fluctuation threshold, identifying the resource conflict nodes associated with the risky process, and determining the interference propagation path, includes: Extract the predetermined operation time window from the preliminary production schedule and obtain the historical operation data of the corresponding equipment for the process; Calculate the standard deviation or range of the historical operating data of the equipment relative to the predetermined operation time window, and determine it as the process time fluctuation index; If the process time fluctuation index is greater than the preset fluctuation threshold, the corresponding process will be marked as the risk process. Identify the equipment nodes and material nodes occupied by the risky process, determine the subsequent affected process sequence, and generate the interference propagation path.

5. The intelligent scheduling method for continuous multi-process production of automotive structural components according to claim 1, characterized in that, The step of calculating the buffer time margin of the interference propagation path and generating an adjustment demand signal containing a rearrangement priority based on the proportion of the buffer time margin consumed includes: Obtain the reserved buffer time for each process along the interference propagation path, and calculate the total buffer time margin; The estimated delay time is calculated based on the process time fluctuation index of the risk process, and the proportion of the estimated delay time to the total buffer time is calculated to obtain the consumed proportion. If the consumption ratio exceeds the preset high-intensity threshold, a first-level response signal is generated, and the process on the interference propagation path is identified as the first priority reordering object. If the consumption ratio is within the preset medium intensity range, a secondary response signal is generated, and the process on the interference propagation path is identified as the second priority rearrangement object. The priority marker containing the first-level response signal or the second-level response signal is encapsulated as the adjustment requirement signal.

6. The intelligent scheduling method for continuous multi-process production of automotive structural components according to claim 1, characterized in that, The initial temperature parameters of the simulated annealing algorithm are set according to the rearrangement priority in the adjustment demand signal. Based on the initial temperature parameters, a neighborhood transformation is performed on the preliminary production schedule, and an iterative search is conducted using a preset multi-objective energy function to obtain an optimized production sequence, including: If the rearrangement priority is the first priority, then the initial temperature parameter is set to a preset high temperature value, and the neighborhood exchange operation is determined to be the corresponding neighborhood transformation method. If the rearrangement priority is the second priority, then the initial temperature parameter is set to a preset medium temperature value, and the insertion neighborhood operation is determined to be the corresponding neighborhood transformation method. Construct the multi-objective energy function, which includes a device idle time term, a cumulative process delay term, and a production rhythm deviation term; Candidate sequences are generated based on the initial temperature parameters and the neighborhood transformation method. The system energy value of the candidate sequences is calculated, and the optimal solution is updated according to the preset acceptance probability criterion until the preset iteration termination condition is met, and the optimized production sequence is output.

7. The intelligent scheduling method for continuous multi-process production of automotive structural components according to claim 1, characterized in that, The process of calculating the production disruption factor by combining the idle process time window with the material availability status, and embedding the emergency order response demand into the optimized production sequence based on the production disruption factor to determine the final execution plan includes: Identify the time periods in the optimized production sequence where no tasks are assigned, and determine them as the idle process time windows; Obtain the list of materials required to respond to the emergency order request, and check the current inventory status of the list of materials to obtain the material availability status; The delay time of subsequent processes caused by embedding the emergency order response demand into each idle process time window is calculated and determined as the production disturbance factor. The time window with the smallest production disturbance factor is selected as the minimum disturbance insertion point; The emergency order response requirement is embedded into the minimum disturbance insertion point, and subsequent processes are sequentially reorganized to generate the final execution plan.

8. The intelligent scheduling method for continuous multi-process production of automotive structural components according to claim 3, characterized in that, The step of loading the final execution plan into a preset discrete event simulation engine to generate a virtual job trajectory, and comparing the virtual job trajectory with the sensor data stream to generate deviation correction parameters, includes: Run the discrete event simulation engine and output the virtual job trajectory containing the simulated completion time; Collect actual processing progress data on site, calculate the difference between the simulated completion time and the actual completion time, and generate an expected deviation dataset; Analyzing the expected deviation dataset, if the deviation shows a continuous cumulative trend, it is extracted as a systematic drift feature; if the deviation shows an occasional distribution, it is extracted as a random drift feature; the systematic drift feature and the random drift feature are collectively referred to as hysteresis drift features. For the systematic drift characteristics, a buffer time weight coefficient for the corresponding process is generated; for the random drift characteristics, a priority correction coefficient for the corresponding process is generated; the buffer time weight coefficient and the priority correction coefficient are combined to form the deviation correction parameter.

9. The intelligent scheduling method for continuous multi-process production of automotive structural components according to claim 8, characterized in that, After parsing the final execution plan to generate the production instruction sequence, the process further includes: Based on the deviation correction parameters, adjust the equipment occupancy weight and material priority weight of different process types in the preset scheduling rule base; The adjusted weights are written into the scheduling rule base, and the scheduling rule base is marked as the benchmark parameter for calculating the fusion feature matrix or generating the initial population in the next production cycle. The adjusted scheduling rule base is periodically verified. If the hysteresis drift characteristic is lower than the preset stability threshold for a consecutive preset period, the current weight is fixed.

10. An intelligent scheduling system for continuous multi-process production of automotive structural components, characterized in that, include: The data fusion module is used to acquire equipment status data and material flow location data for each process in the production line, and to preprocess the data to obtain the current production status vector. The initial production scheduling module is used to iteratively solve multi-process tasks using a genetic algorithm based on the current production state vector to obtain a preliminary production schedule. The interference detection module is used to obtain the process time fluctuation index in the preliminary production schedule. If the process time fluctuation index exceeds the preset fluctuation threshold, the risk process is marked, the resource conflict node associated with the risk process is identified, and the interference propagation path is determined. The buffer time margin of the interference propagation path is calculated, and an adjustment demand signal containing the reordering priority is generated according to the proportion of the buffer time margin consumed. The dynamic optimization module is used to set the initial temperature parameters of the simulated annealing algorithm according to the rearrangement priority in the adjustment demand signal; perform neighborhood transformation on the preliminary production schedule based on the initial temperature parameters; and perform iterative search using a preset multi-objective energy function to obtain an optimized production sequence. The order insertion response module is used to obtain urgent order insertion response requirements, extract idle process time windows from the optimized production sequence, and obtain the current material availability status; calculate the production disturbance factor by combining the idle process time windows and the material availability status, and embed the urgent order insertion response requirements into the optimized production sequence based on the production disturbance factor to determine the final execution plan; The feedback correction module is used to acquire real-time sensor data streams collected on-site. The final execution plan is loaded into a preset discrete event simulation engine to generate a virtual job trajectory, and the virtual job trajectory is compared with the sensor data stream to generate deviation correction parameters. The final execution plan is parsed to generate a production instruction sequence.

Images