An artificial intelligence-based automated production line optimization method

Abstract

This invention discloses an artificial intelligence-based optimization method for automated production lines, belonging to the field of production line automation technology. The method includes: collecting process topology relationships, equipment status data, material inventory data, and order delivery date data of the production line; constructing a process-equipment capacity matching graph, mapping each process to a currently idle equipment node with processing capacity; calculating the dynamic priority index of each process based on order delivery urgency and process duration; identifying process blocking nodes caused by equipment conflicts, adjusting the execution order of blocking processes using a backtracking conflict resolution strategy, and generating a conflict-free process sequence; performing feasibility verification on the process sequence based on material kitting constraints, postponing processes that do not meet the kitting conditions to the earliest available time window after all materials are available; and outputting the optimized process scheduling scheme and sending it to the production line execution system. This invention achieves multi-constraint collaborative optimization, improving the feasibility of process scheduling and production efficiency.

Description

Technical Field

[0001] This invention relates to the field of production line automation technology, specifically to an artificial intelligence-based method for optimizing automated production lines. Background Technology

[0002] In automated production line scheduling, the multi-variety, variable-batch production mode leads to complex inter-process dependencies, tight equipment resources, and significant uncertainty in material supply. Traditional scheduling methods often rely on static priority rules, such as first-come-first-served or shortest processing time priority, which can generate usable scheduling solutions in scenarios with a single product or sufficient equipment.

[0003] However, existing technologies typically handle equipment conflicts by simply delaying or preemptively adjusting, without comprehensively assessing the cascading effects of the conflict. This can easily lead to widespread delays in subsequent processes, and the adjustment results are limited by the local decision-making sequence, making it difficult to resolve the blockages caused by competition for equipment resources. Regarding material constraints, existing scheduling processes often perform independent material availability checks after generating the process sequence. Material shortage information cannot be fed forward to the scheduling construction stage, resulting in a disconnect between the scheduling plan and the actual material arrival status. This leads to frequent downtime due to material shortages or forced order insertions, increasing production line disruptions.

[0004] Therefore, it is necessary to address how to effectively resolve process blockages caused by competition for equipment resources during the scheduling process without introducing new bottlenecks, and how to dynamically integrate material kitting constraints into the process sequence generation process so that the scheduling results remain executable under material supply fluctuations. Summary of the Invention

[0005] This paper presents an AI-based optimization method for automated production lines. During the scheduling process, a backtracking conflict resolution strategy is used to eliminate process blockages caused by equipment conflicts. At the same time, material kitting verification is embedded in the process sequence generation process, so that the output solution has higher executability while meeting equipment capacity constraints and material supply constraints.

[0006] To achieve the above objectives, this invention provides the following technical solution: This invention provides an automated production line optimization method based on artificial intelligence. The method includes: collecting process topology relationships, equipment status data, material inventory data, and order delivery date data of the production line, thereby forming a comprehensive perception of the dynamics of the production site. A process equipment capacity matching graph is constructed, mapping each process to a currently idle equipment node with processing capacity, achieving precise coupling between processes and equipment resources. Based on the urgency of order delivery dates and process consumption time, a dynamic priority index for each process is calculated, quantifying the urgency of the process. Process blocking nodes caused by equipment conflicts are identified, and a backtracking conflict resolution strategy is used to adjust the execution order of blocking processes, generating a conflict-free process sequence, thereby effectively resolving resource contention. The feasibility of the process sequence is verified based on material availability constraints, postponing processes that do not meet the material availability conditions to the earliest available time window after material availability, ensuring real-time matching between process scheduling and material supply. The optimized process scheduling scheme is output and sent to the production line execution system to guide production operations.

[0007] In constructing the process-equipment capacity matching graph, all processes are traversed, and equipment with corresponding processing capabilities is selected from the equipment database to generate a candidate pairing set of processes and equipment. Each equipment node is marked with its currently occupied time slot and remaining available time. Edges are established between process nodes and candidate equipment nodes, with the edge weight assigned to the estimated processing time of that process on that equipment. A maximum weight matching algorithm is used to select a unique equipment for each process from the candidate pairing set, forming a matching graph with no duplicate equipment allocation. This process ensures precise alignment between equipment capacity and process requirements in the time dimension, maximizing equipment utilization.

[0008] When calculating the dynamic priority index for each process, the remaining grace time is calculated by extracting the difference between the remaining delivery date of the order to which each process belongs and the current time. The number of subsequent processes dependent on this process and their total processing time are then calculated to obtain the path urgency of the process. The reciprocal of the remaining grace time is multiplied by the path urgency, and then multiplied by the material shortage risk coefficient for this process to obtain the initial priority value. The initial priority value is then normalized, outputting a dynamic priority index between 0 and 1. This index accurately reflects the urgency of the process in the order delivery chain and the material availability risk, providing a ranking basis for subsequent conflict resolution. Preferably, the material shortage risk coefficient is dynamically calculated based on the historical supply volatility of the material corresponding to the process and the arrival confidence of the current in-transit purchase orders, making risk prediction closer to the actual supply situation.

[0009] When identifying process blockage nodes caused by equipment conflicts, the current process sequence is traversed, and for each process, it is checked whether its matching equipment is already occupied by other processes within the planned execution time window. If occupied, the process is marked as a blockage node, and the conflicting equipment identifier and conflict time interval are recorded. The waiting margin of the blockage node is calculated, which is the difference between the latest start time that the process can be postponed and the current planned start time. When the waiting margin is less than a preset threshold, the blockage node and its subsequent dependent processes are marked as a set to be adjusted. This margin-based blockage identification mechanism can accurately capture process groups that cannot resolve conflicts by waiting on their own, avoiding cascading delays caused by local delays.

[0010] As a technical solution of this invention, adjusting the execution order of blocked processes using a backtracking conflict resolution strategy includes: selecting the process with the lowest dynamic priority index from the processes already scheduled on the conflicting equipment as a removable process; deleting the removable process from its original equipment queue and placing it in a reassignment pool; re-executing the process equipment capacity matching step for the removable process to allocate it to another idle equipment or postpone it to the next idle time slot of that equipment; if no feasible allocation can be found, backtracking to the previous conflicting node and recursively executing the removal and reassignment operation until all conflicts are eliminated. By combining backtracking and reassignment, a better process arrangement can be searched globally, rather than just performing local insertions, significantly improving the overall feasibility of the process sequence and the equipment load balancing.

[0011] When performing feasibility verification on a process sequence based on material availability constraints, a bill of materials is created for each process, including the type, specifications, and quantity of required raw materials; the real-time inventory of the current material inventory table is queried, and the remaining quantity after adding all materials consumed before that process is calculated; if the remaining quantity of any material required by the target process is less than the demand, the process is determined not to meet the material availability condition; the expected replenishment time of the missing material is recorded, and the planned start time of the process is shifted to the earliest time slot after the replenishment time. By incorporating the time-series cumulative effect of material consumption into the availability judgment, the hidden material shortage risk caused by batch process consumption can be exposed early. Preferably, when shifting the execution plan time, the supplier confirmation delivery time and inventory safety lead time of the missing material are obtained, and the latest available time is calculated; based on the latest available time, the first idle time slot of the candidate equipment for that process after that time is found in the process equipment capacity matching diagram; the planned start time of that process is adjusted to the start time of the idle time slot, and the start time of its subsequent dependent processes is updated synchronously, using topological order to postpone sequentially. This approach precisely matches the certainty of material arrival with equipment availability windows, reducing wasted capacity caused by blindly pushing back production. Furthermore, the material availability condition also includes compatibility checks on substitute materials. When the primary material is out of stock, if a substitute material meets the same specifications and has sufficient quantity, the process is still considered to meet the availability condition, thereby improving the resilience of the scheduling scheme and its adaptability to supply fluctuations. During the sequential postponement of the topology, if the postponement of any subsequent dependent process causes its planned completion time to exceed the order delivery date, an emergency order insertion warning is triggered, and a retrospective conflict resolution strategy is re-executed. This forms a closed-loop adjustment between material constraints and equipment constraints, ensuring that the final scheduling scheme maximizes the fulfillment of delivery dates.

[0012] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0013] To resolve equipment conflicts, a backtracking conflict resolution strategy is adopted: When traversing the process sequence, if equipment is detected to be occupied by other processes within the planned time window, that process is marked as a blocked node. From the processes already scheduled on the conflicting equipment, the process with the lowest dynamic priority index is selected as a removable process, removed from the original equipment queue, and placed in the reassignment pool. The removable process undergoes a new process-equipment capacity matching process, allocating it to another idle equipment or postponing it to the next idle time slot of that equipment. If no feasible allocation is found, the process backtracks to the previous conflicting node, recursively performing the removal and reassignment operation until all conflicts are resolved. This strategy, guided by the dynamic priority index, prioritizes ensuring the occupation of critical equipment by high-urgency processes. Through cross-equipment replanning or time slot postponement of removable processes, it recursively resolves equipment conflicts under the constraint of process topology dependencies, expanding the feasible search space, avoiding the accumulation of blockages caused by local postponement, and improving the globality of process-equipment matching.

[0014] To dynamically incorporate material availability constraints, a material shortage risk coefficient is introduced when calculating the dynamic priority index for each process. This coefficient is dynamically calculated based on the historical supply volatility of the corresponding material for that process and the arrival confidence of current in-transit purchase orders, giving processes with high material supply uncertainty higher scheduling priority. In the material availability feasibility verification, a bill of materials is created for each process, and current inventory and pre-consumption are queried. If the remaining quantity of any material is less than the demand, the estimated replenishment time for the missing material is recorded. Using the latest available time as a benchmark, the first idle time slot of candidate equipment is found in the process equipment capacity matching diagram, and the planned start time of the process is adjusted to that time slot. Subsequent dependent processes are then postponed sequentially using topological order. If the postponement of any subsequent process causes its planned completion time to exceed the order delivery date, an emergency order insertion warning is triggered, and a retrospective conflict resolution strategy is re-executed. This mechanism feeds forward dynamic material supply information to priority calculation and scheduling time window determination, automatically postponing processes with material shortages until material availability is achieved, and synchronously propagating time offsets to form an executable material scheduling linkage scheme, reducing plan failures caused by material mismatches. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0016] Figure 1 This is a flowchart of an AI-based automated production line optimization method. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0018] See Figure 1This invention provides an AI-based method for optimizing automated production lines, comprising: collecting process topology relationships, equipment status data, material inventory data, and order delivery date data of the production line; constructing a process equipment capacity matching graph, mapping each process to a currently idle equipment node with processing capacity; calculating the dynamic priority index of each process based on the urgency of order delivery and process consumption time; identifying process blocking nodes caused by equipment conflicts, adjusting the execution order of blocking processes using a backtracking conflict resolution strategy, and generating a conflict-free process sequence; performing feasibility verification on the process sequence based on material availability constraints, postponing processes that do not meet the material availability conditions to the earliest available time window after the materials are available; outputting the optimized process scheduling scheme and sending it to the production line execution system.

[0019] Example 1

[0020] In practice, the process of constructing the process-equipment capacity matching map begins with obtaining all processes to be scheduled on the production line. The automated production line optimization system reads the process topology data and extracts the identifier, required processing capacity type, and technical requirements parameters for each process. The equipment database stores the identifier, processing capacity set, current operating status, and allocated time slot information for each piece of equipment. The system traverses all processes. For each process encountered, it queries the equipment database based on the required processing capacity type, filtering out equipment whose processing capacity set contains the required processing capacity type and whose current status is idle or about to become idle. This generates a candidate pairing set consisting of process identifiers and equipment identifiers. Each element in the candidate pairing set is a two-dimensional tuple, where the first dimension is the process identifier and the second dimension is the equipment identifier capable of processing that process.

[0021] For each device node corresponding to a device identifier appearing in the candidate pairing set, the system retrieves a list of allocated time slots for that device node from the device database. A time slot is defined by a start time and an end time, representing the period during which the device is occupied. The system marks all occupied time slots for the device node on the timeline and calculates the remaining available time for the device node by subtracting the sum of the durations of the occupied time slots from the total available time of the device node. The remaining available time is calculated as follows: Let the total available time of the device node for the day be... The time slots already occupied are One, of which the first The duration of each time slot is The expression for calculating the remaining available time is:

[0022]

[0023] in, Indicates the remaining available time of the device node. This represents the total available time of a device node within the scheduling period. This total available time is determined based on the device node's scheduling and maintenance plans. This indicates the number of time slots allocated to the device node. Indicates the device node number The duration of an occupied time slot.

[0024] After generating the candidate pairing set and marking the time slots of the equipment nodes, the system establishes an edge between the process node and the equipment node in each candidate pairing. The conditions for establishing an edge are: the processing capacity type required by the process represented by the process node matches the processing capacity type of the equipment represented by the equipment node, and the equipment node has a continuous idle period of not less than the estimated processing time within the planned execution time interval of the process. The weight of the edge is assigned as the estimated processing time of the process represented by the process node on the equipment represented by the equipment node. The estimated processing time is obtained as follows: the system extracts processing time samples of the same process type on the same or similar equipment from historical processing records, removes outliers, and takes the average to obtain the baseline processing time; then, it corrects the baseline processing time based on the current operating status factors of the equipment node, including the wear compensation coefficient and the current load coefficient of the equipment node. The corrected time is used as the weight of the edge.

[0025] After assigning edge and weight values, the system uses a maximum weight matching algorithm to select a unique device for each process from the candidate pairing set. The core architecture of the maximum weight matching algorithm is as follows: a bipartite graph is constructed by connecting the set of process nodes and the set of device nodes connected to the process nodes with edges. The left side of the bipartite graph contains all process nodes, and the right side contains all relevant device nodes. The edges and weights have been given in the preceding steps. The maximum weight matching algorithm is implemented using the Kuhn-Munkres algorithm. This algorithm obtains the maximum weight matching result of the bipartite graph through an iterative process of initializing feasible vertex labels, constructing an equality subgraph, finding augmenting paths, and updating vertex labels. During the iteration process, the initial value of the vertex label of the left-hand process node is set to the maximum value among the weights of all edges connected to that process node, and the initial value of the vertex label of the right-hand device node is set to zero. The equality subgraph consists of edges that satisfy the condition that the sum of the vertex labels of the process nodes and the device nodes equals the edge weight. The algorithm searches for augmenting paths for each unmatched process node in the equality subgraph. If an augmenting path is found, the matching relationship is updated; if no augmenting path is found, the top labels of the involved process nodes and equipment nodes are adjusted to expand the equality subgraph, and the search for augmenting paths continues until all process nodes are matched or the equality subgraph cannot be expanded further. In the final matching result, each process node uniquely corresponds to one equipment node, and each equipment node is assigned at most one process within the same time slot, forming a matching graph with no duplicate equipment assignments. Each matching pair in the matching graph indicates the arrangement relationship of the process on the selected equipment, and subsequent process sequence scheduling will be based on this matching graph.

[0026] Example 2

[0027] In practice, the calculation of the dynamic priority index for each process begins with extracting the remaining grace time for the orders to which each process belongs. The automated production line optimization system obtains the order delivery date of each order to which each process belongs from the order delivery date data and obtains the current time from the system clock. For each process traversed, the system subtracts the current time from the order delivery date of the order to which the process belongs to obtain the remaining grace time. The remaining grace time is measured in minutes or hours and is used to characterize the degree of time leeway for the process while meeting the order delivery date. The smaller the remaining grace time, the higher the time urgency of the process.

[0028] After obtaining the remaining grace time, the system calculates the path urgency of the process. Path urgency is determined by the number of subsequent processes that the current process depends on, and the total processing time of these dependent processes. The system extracts all subsequent processes starting from the current process from the process topology data. By performing a depth-first traversal along the edges of the process topology, it counts the number of all reachable subsequent processes and sums the estimated processing time of each reachable subsequent process on standard equipment to obtain the total processing time. The more subsequent processes a process depends on and the longer the total processing time, the greater the impact of a delay in this process on the entire production chain, and the higher the path urgency. The path urgency is calculated by multiplying the number of subsequent processes that the current process depends on by the total processing time.

[0029] The system dynamically calculates the material shortage risk coefficient based on the historical supply volatility of materials corresponding to each process and the arrival confidence of current purchase orders in transit. The material shortage risk coefficient is obtained by retrieving the historical supply volatility of each material required for the process from the material management database. The deviation between the actual and planned delivery times within each supply cycle is recorded to calculate historical supply volatility. The historical supply volatility is calculated as follows: for each material, [the following data is used for calculation]. The absolute value of the difference between the actual and planned delivery times within each supply cycle, divided by the planned supply interval, is then used to calculate the actual delivery time. The historical supply volatility of the material is obtained by averaging the calculation results. The value is That is, take the past Data for each supply cycle, The basis for setting each supply cycle is based on supply chain management practices. Each cycle can cover seasonal fluctuations and recent trend changes, while avoiding misjudgments of current supply conditions based on overly distant data. The system also extracts the estimated arrival times of current in-transit purchase orders from the procurement system, using the supplier's historical on-time delivery rate as the arrival confidence level. For each material, the system correlates the material's historical supply volatility with (… The material shortage risk coefficient is obtained by multiplying the following factors (subtracting the arrival confidence level) and summing them, then dividing by the required quantity and type of materials for the process. The value range of the material shortage risk coefficient is within [a certain range]. arrive Between these values, a larger value indicates a higher risk of material shortage.

[0030] The system takes the reciprocal of the remaining slack time, multiplies this reciprocal by the path urgency, and then multiplies by the material shortage risk coefficient to obtain the initial priority value. The expression for calculating the initial priority value is as follows:

[0031]

[0032] in, This indicates the initial priority value of the process. This indicates the remaining grace time for the order to which the process belongs. This indicates the path urgency of a process. Path urgency is calculated by multiplying the number of subsequent processes that the process depends on by the total processing time of those subsequent processes. This represents the material shortage risk coefficient for the process, and the value range of the material shortage risk coefficient is within... arrive The material shortage risk coefficient is dynamically calculated based on the historical supply volatility of the corresponding materials for the process and the arrival confidence of the current in-transit purchase orders.

[0033] After obtaining the initial priority value, the system normalizes it. The normalization method is as follows: Obtain the initial priority values ​​of all processes in the current batch, find the maximum and minimum values, and use the min-max normalization method. Subtract the minimum value from the initial priority value of each process, and then divide the result by the difference between the maximum and minimum values ​​to obtain the normalized value. When the normalized value is... When the time is specified, it indicates that the process has the lowest priority in the current batch; the normalized value is... When the time is specified, it indicates the highest priority. The output after normalization is between [time range missing]. and The dynamic priority index between them.

[0034] The dynamic priority index is used in subsequent process blockage identification and conflict resolution. In the backtracking conflict resolution strategy, when it is necessary to select a removable process from multiple blocked processes, the system selects based on the dynamic priority index, with processes having lower dynamic priority indices being prioritized as removable processes.

[0035] Example 3

[0036] In practice, the process of identifying process blockage nodes caused by equipment conflicts begins by traversing the current process sequence. The automated production line optimization system obtains the current process sequence after allocation through the process equipment capacity matching map. Each process in the current process sequence is bound to a equipment node selected by the maximum weight matching algorithm and a planned execution time window determined based on the equipment idle time slot. The system performs equipment occupancy conflict checks on each process in the current process sequence in order.

[0037] For each process traversed in the current process sequence, the system obtains the identifier of the matching equipment node in the process equipment capacity matching diagram and the planned execution time window for that process. The planned execution time window is defined by the planned start time and the planned end time, with the planned end time being the planned start time plus the estimated processing time of the process on the matching equipment. The system queries the time slot occupancy record of the matching equipment node, which stores all scheduled processes on that equipment node and their start and end times. The system compares the planned execution time window of the currently processed process with the time slots occupied by all scheduled processes on the matching equipment node one by one. If there is an overlap between the time slot occupied by a scheduled process and the planned execution time window of the currently processed process on the time axis, it is determined that there is a equipment conflict for the currently processed process on the matching equipment node.

[0038] When a device conflict is detected, the system marks the currently processed operation as a blocked node and generates a blocking record. The blocking record includes the operation identifier of the blocked node, the identifier of the conflicting device, and the conflict time interval. The conflict time interval is the start and end time of the overlapping portion between the planned execution time window of the currently processed operation and the time slot occupied by the scheduled operation.

[0039] After marking a blocked node, the system calculates its waiting margin. The waiting margin is defined as the difference between the latest deferable start time of the blocked node and its currently planned start time. The latest deferable start time is determined as follows: extract the order delivery date of the order to which the blocked node belongs; obtain the estimated processing time of the blocked node on the matching equipment; extract all subsequent dependent processes starting from the blocked node from the process topology data; and obtain the sum of the estimated processing times of each subsequent dependent process on its corresponding equipment; subtract the estimated processing time of the blocked node from the order delivery date, and then subtract the sum of the estimated processing times of all subsequent dependent processes to obtain the latest deferable start time. The expression for calculating the waiting margin is:

[0040]

[0041] in, This represents the waiting margin of the blocked node. This indicates the order delivery time of the order to which the blocked node belongs. This indicates the estimated processing time of the blocked node on the matching device. This represents the total number of subsequent dependent processes of the blocked node. The first node representing the blocked node The estimated processing time for each subsequent dependent process. Indicates the current planned start time of the blocked node.

[0042] The system obtains a preset threshold and compares the calculated waiting margin with the preset threshold. The preset threshold is set to... Minutes are set based on the following: when the waiting margin is less than... At a certain time, it indicates that even if the start time of the blocked node is postponed to the latest possible start time, it is still impossible to complete the processing of the blocked node and its subsequent dependent processes before the order delivery time. This blocked node does not have the time margin to resolve the conflict through simple delay and must be included in the conflict resolution adjustment scope. When the waiting margin is less than a preset threshold... At the specified minute mark, the system marks the blocked node and all its subsequent dependent processes, found along the dependency direction using the process topology data, as a set to be adjusted. All processes within a set to be adjusted are treated as a single adjustment unit during subsequent conflict resolution.

[0043] After identifying the blocking nodes in the process and marking the set to be adjusted, the system uses a backtracking conflict resolution strategy to adjust the execution order of the blocking processes.

[0044] In practice, the backtracking conflict resolution strategy takes as input the set to be adjusted and outputs a conflict-free sequence of work processes. The core architecture of the backtracking conflict resolution strategy is a recursive solution module, which consists of three layers: a conflict selection layer, a removal and reassignment layer, and a backtracking control layer. The conflict selection layer is responsible for selecting removable work processes from the already scheduled work processes on conflicting equipment; the removal and reassignment layer is responsible for removing removable work processes from the original equipment queue and finding new feasible assignments for them; the backtracking control layer is responsible for triggering a backtracking operation when no feasible assignment can be found, reverting to the previous conflict node to reselect and reassign.

[0045] The conflict selection layer process is as follows: For each conflicting device corresponding to a blocked node in the set to be adjusted, the system obtains a list of all currently scheduled operations on the conflicting device. Operations overlapping with the planned execution time windows of the blocked nodes are selected from this list, forming a conflict operation set. The system reads the dynamic priority index of each conflicting operation in the conflict operation set; the value of the dynamic priority index has been pre-calculated and stored during this scheduling process. The system selects the operation with the smallest dynamic priority index from the conflict operation set as the removable operation.

[0046] The process of removing and reallocating the process layer is as follows: The system deletes the removable process from the time slot occupancy record of the conflicting equipment, releasing the time slots originally occupied by the removable process on the conflicting equipment. The system then places the removable process into the waiting-to-relocate pool. The waiting-to-relocate pool is a cached list that stores processes that have been removed and are waiting to be rematched. The system re-executes the process equipment capability matching step for the removable processes in the waiting-to-relocate pool. When re-executing the process equipment capability matching step, the system uses the removable process as input, filters out equipment with corresponding processing capabilities from the equipment database, excludes equipment nodes that are currently marked as unavailable, and generates a candidate pairing set; it marks the currently occupied time slots and remaining available time for the candidate equipment nodes; then it allocates an idle equipment node different from the original conflicting equipment to the removable process, or finds the next consecutive idle time slot on the original conflicting equipment to which the removable process belonged, the length of which is not less than the estimated processing time of the removable process. If a suitable allocation is found, the system will assign the removable process to the idle time slot of the selected device node, update the time slot occupancy record of the device node, and remove the removable process from the pool to be reassigned, thus resolving the conflict of the blocking node.

[0047] The backtracking control layer processes as follows: If the removal and redistribution layer cannot find a feasible allocation for the removable operation, meaning that all equipment nodes with processing capabilities have no available time slots within an acceptable time range, the backtracking control layer triggers a backtracking operation. The backtracking operation marks the currently blocked node as unresolved and cancels the removal operation for the removable operation, restoring the removable operation to the original time slot of the conflicting equipment. The system backtracks to the previous conflicting node, which is maintained by the conflict resolution stack. Each time a removable operation is selected, the conflict resolution stack pushes the conflicting node information onto the top of the stack. This information includes the conflicting equipment identifier, the set of conflicting operations, and a list of operations that have been attempted to be removed. During backtracking, the system pops the information of the previous conflicting node from the top of the conflict resolution stack, excludes previously attempted removal operations from the conflicting operation set of the previous conflicting node, and reselects the operation with the second lowest dynamic priority index as the new removable operation, recursively executing the removal and redistribution operation. The recursive execution terminates when all conflicting nodes have been resolved and the set to be adjusted is empty; at this point, the system outputs the adjusted sequence of conflict-free operations.

[0048] Example 4

[0049] In practice, the feasibility verification of the process sequence based on material kitting constraints begins with the creation of a bill of materials (BOM) for each process. The automated production line optimization system extracts the material codes, specifications, and quantities of raw materials required for each process from the product formula database and process route data, establishing a correspondence table between processes and material requirements details. This correspondence table constitutes the BOM for each process. In the BOM, each material corresponds to a material code, which is used to uniquely identify the material in the material inventory table. The material specifications include the material's model, size, and material attributes, used for specification comparison in subsequent compatibility checks with alternative materials. The material quantity represents the quota amount of the material consumed in a single execution of the process.

[0050] After obtaining the bill of materials (BOM) for each process, the system performs material kitting checks on each process in the process sequence according to the scheduling order. For the target process currently being checked, the system queries the real-time inventory of the current material inventory table. The material inventory table is updated in real time by the production line warehouse management system, recording the current actual quantity of each material in stock. The system performs inventory balance calculation for each material in the BOM of the target process. The calculation method is as follows: the consumption of the material in all scheduled and material kitting checks of the preceding processes before the target process is introduced. The consumption of the preceding processes is obtained by adding the demand of the same material in the BOMs of all preceding processes in the process sequence of the target process. The estimated remaining quantity of the material before the execution of the target process is obtained by subtracting the total consumption of the material in all preceding processes from the real-time inventory of the material inventory table. The expression for calculating the estimated remaining quantity is:

[0051]

[0052] in, The material code before the target process is executed is: The estimated remaining quantity of materials; The material code is The real-time inventory of the materials in the current material inventory table; This represents the total number of preceding processes in the process sequence that are located before the target process and have passed the material kitting check. Indicates the first The material code in the bill of materials for each preceding process is: The required quantity of materials. When the target process is the first process in the process sequence. Values ,at this time for The estimated remaining amount is equal to the real-time inventory.

[0053] The system compares the required quantity of each material in the bill of materials for the target process with its corresponding estimated remaining quantity. If the required quantity of any material exceeds its estimated remaining quantity, the target process is deemed not to meet the material availability criteria. For target processes determined not to meet the material availability criteria, the system extracts the estimated replenishment time for the missing materials from the procurement management system or supply chain collaboration platform. The estimated replenishment time is derived from the supplier's confirmed delivery time for the purchase order combined with transportation timeliness data. In the process equipment capacity matching diagram, using the estimated replenishment time as a benchmark, the system searches for the earliest available time slot after the estimated replenishment time in the idle time slots of candidate equipment for the target process, and shifts the planned start time of the target process to the start time of that earliest available time slot. After the shift operation is completed, the system marks the target process as having passed the material availability check and updates the time slot occupancy record of the corresponding equipment for the target process.

[0054] During the material availability determination process, the system simultaneously performs a compatibility check on substitute materials. The implementation method for substitute material compatibility checking is as follows: when the expected remaining quantity of any primary material in the target process's bill of materials is detected to be less than the required quantity, the system does not immediately determine that the material availability condition is not met. Instead, it queries the substitute material relationship database. The substitute material relationship database pre-stores the correspondence between each primary material and the substitute material that can be used to replace it, as well as the equivalence description of the substitute material and the primary material in terms of specifications. The system searches the substitute material relationship database to see if there is a substitute material record for the primary material that is in short supply. If a substitute material record exists, the system obtains the material code of the substitute material and queries the real-time inventory of the substitute material in the current material inventory table. Simultaneously, referring to the primary material consumption calculation method, it calculates the expected remaining quantity of the substitute material before the target process is executed. If the expected remaining quantity of the substitute material is greater than or equal to the required quantity of the primary material, the system determines that the target process still meets the material availability condition and marks the substitute material as the material consumed in this process in the material consumption record. The consumption amount is calculated by deducting the inventory of the substitute material from the required quantity of the primary material. If the estimated remaining quantity of substitute materials is also insufficient, the system will continue to maintain the judgment that the target process does not meet the material availability conditions, and will execute the recording of the estimated replenishment time of the missing materials and the operation of postponing the planned start time.

[0055] Example 5

[0056] In practice, the process of shifting the planned start time of a process begins with confirming the delivery time and safety margin of the missing materials from the supplier. During the feasibility verification of the process sequence based on material availability constraints, the automated production line optimization system initiates the shift operation when it determines that the target process does not meet the material availability conditions and records the expected replenishment time of the missing materials.

[0057] The system retrieves the supplier-confirmed arrival time for the materials in short supply from the procurement management system. The supplier-confirmed arrival time is the moment, accurate to the minute, promised by the supplier in the purchase order confirmation receipt when the goods will arrive at the production line warehouse. The system also reads the inventory safety lead time from the inventory management configuration parameters. The inventory safety lead time is defined as a buffer period reserved to accommodate supplier delivery delays or time-consuming material inspection upon receipt. The inventory safety lead time is obtained by querying the material management database for historical arrival delay records of the materials in short supply. The average delay between the actual delivery time and the supplier's confirmed delivery time for each shipment. Values Set as The basis is that in statistics, the threshold for large samples is usually taken as... However, considering that some materials are purchased infrequently, we will select [a certain number of items] while ensuring the representativeness of the sample. This can cover delivery data for nearly six months to a year; the calculated average delay time is added to the material's inbound quality inspection standard time, which is obtained from the quality inspection process documents, and the sum is used as the inventory safety lead time.

[0058] The system adds the supplier's confirmed delivery time to the inventory safety lead time to calculate the latest available time. The expression for calculating the latest available time is:

[0059]

[0060] in, Indicates the latest time when the material in short supply will be available. This indicates the supplier's confirmed delivery time for the materials that are in short supply. This indicates the lead time for safe inventory, which is calculated by adding the historical average delay time of the arrival of the missing materials to the time taken for the quality inspection standards upon receipt.

[0061] After calculating the latest available time, the system uses this latest available time as a benchmark to find the first idle time slot for the candidate equipment of the target process after that latest available time in the process equipment capacity matching graph. The search process is as follows: The system obtains a list of candidate equipment for the target process from the process equipment capacity matching graph. The candidate equipment list contains all equipment nodes that have the capability to process the target process and are currently marked as available. For each candidate equipment in the candidate equipment list, the system obtains the time slot occupancy record for that candidate equipment. The time slot occupancy record contains the start and end times of the occupancy of all scheduled processes on the candidate equipment, as well as the idle gap between any two adjacent occupied time slots. Starting from the latest available time point on the candidate equipment's time axis, the system scans the time slot occupancy record backwards to find the first consecutive idle time period that meets the following conditions: the start time of the idle time period is not earlier than the latest available time, and the total duration of the idle time period is not less than the estimated processing time of the target process on that candidate equipment. The system uses the start time of the found idle time period as the candidate start time of the target process on that candidate equipment. The system iterates through all candidate devices in the candidate device list, obtains a candidate start time for each candidate device, and selects the earliest start time from all candidate start times as the adjusted planned start time for the target process. If multiple candidate devices have the same earliest start time, the candidate device with the shortest estimated processing time is selected as the device to be assigned to the target process.

[0062] The system adjusts the planned start time of the target process to the start time of the selected idle time slot, updates the time slot occupancy record of the target process in the process equipment capacity matching diagram, binds the target process to the selected candidate equipment, and occupies the time period between the start time of the selected idle time slot and the start time plus the estimated processing time.

[0063] After adjusting the planned start time of the target process, the system synchronously updates the start times of all subsequent dependent processes. This synchronization is implemented using a topological ordering approach. The system extracts a directed dependency graph from the process topology data, starting with the target process, and performs topological sorting on this graph to obtain an ordered list of subsequent dependent processes. The topological sorting algorithm is implemented using Kahn's algorithm, whose core architecture includes an in-degree calculation module and a zero-in-degree queue processing module. The in-degree calculation module traverses all process nodes in the directed dependency graph, calculating the in-degree value of each node, which is the number of predecessor processes pointing to that node. The zero-in-degree queue processing module adds all process nodes with an in-degree value of zero to a processing queue. It then retrieves process nodes one by one from the head of the queue, outputs them to the ordered list, and decrements the in-degree values ​​of all subsequent process nodes pointed to by that node. If the in-degree value of a subsequent process node becomes zero, it adds that node to the tail of the processing queue. This process is repeated until the processing queue is empty, outputting a complete topologically sorted ordered list.

[0064] The system sequentially postpones the start time of each subsequent dependent operation in the topologically ordered list. For the first subsequent dependent operation in the ordered list, the system adds the estimated processing time of the target operation on the selected equipment to the adjusted planned start time of the target operation, obtaining the constraint start time of the first subsequent dependent operation. The system compares the current planned start time of the first subsequent dependent operation with the constraint start time. If the current planned start time is earlier than the constraint start time, the planned start time of the first subsequent dependent operation is adjusted to the constraint start time, and the entire time slot of the first subsequent dependent operation is shifted backward in the process equipment capacity matching diagram by the difference between the constraint start time and the current planned start time. The shift operation is performed on the matching equipment of the first subsequent dependent operation. If the shifted time slot conflicts with other scheduled operations on the matching equipment, all affected scheduled operations on the matching equipment are cascaded backward according to the same rules. For each subsequent dependent operation in the ordered list that follows, the system uses the adjusted planned completion time of its direct predecessor operation as the constraint start time, repeating the above comparison and shift operations until all subsequent dependent operations in the ordered list have completed their start time updates.

[0065] During the sequential postponement of the topology sequence, the system performs an order delivery date overdue check on each postponed subsequent dependent process. The order delivery date overdue check is implemented as follows: the system obtains the adjusted planned start time of the postponed process, adds the estimated processing time of the postponed process on its matching equipment, and obtains the planned completion time of the postponed process. The system extracts the order delivery date of the order to which the postponed process belongs and compares the planned completion time of the postponed process with the order delivery date. If the postponement of any subsequent dependent process causes its planned completion time to exceed the order delivery date of the order to which it belongs, an emergency order insertion warning is triggered.

[0066] The emergency order insertion warning triggering process involves two steps: generating a warning signal and initiating conflict resolution. The warning signal includes the process identifier that triggered the warning, the associated order identifier, the difference between the planned completion time and the order delivery date, and a complete list of currently affected subsequent dependent processes. The system pushes the warning signal to the production line management terminal, highlighting the out-of-limit process and its associated order information on the management interface. Simultaneously, the system uses the process that triggered the warning and its subsequent dependent processes as a new set to be adjusted, and re-executes the backtracking conflict resolution strategy. When re-executing the backtracking conflict resolution strategy, the input set to be adjusted is the process that triggered the warning and its subsequent dependent processes, and the output is a conflict-free process schedule that, after readjustment, ensures that the planned completion time of all processes does not exceed the order delivery date.

[0067] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes 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.

Claims

1. An automated production line optimization method based on artificial intelligence, characterized in that, include: Collect process topology data, equipment status data, material inventory data, and order delivery date data of the production line; Construct a process equipment capacity matching graph, mapping each process to a device node that has processing capacity and is currently idle; Calculate the dynamic priority index for each process based on the urgency of the order delivery date and the time consumed in each process. Identify process blocking nodes caused by equipment conflicts, and use a backtracking conflict resolution strategy to adjust the execution order of the blocked processes to generate a conflict-free process sequence; Based on the material availability constraints, the feasibility of the process sequence is verified, and processes that do not meet the material availability conditions are postponed to the earliest available time window after the materials are available. The optimized process scheduling plan is output and sent to the production line execution system.

2. The method for optimizing an automated production line based on artificial intelligence according to claim 1, characterized in that, The equipment capability matching diagram for the construction process includes: Iterate through all processes, filter out equipment with corresponding processing capabilities from the equipment database, and generate a candidate pairing set of processes and equipment. For each device node, mark the currently occupied time slot and the remaining available time; Establish an edge between the process node and the candidate device node, and assign the weight of the edge to the estimated processing time of the process on the candidate device; The maximum weight matching algorithm is used to select a unique device for each process from the candidate pairing set, forming a matching graph with no duplicate device allocation.

3. The method for optimizing an automated production line based on artificial intelligence according to claim 1, characterized in that, The calculation of the dynamic priority index for each process includes: Extract the remaining grace time by comparing the remaining delivery date of the order to which each process belongs with the current time; The path urgency of each process is determined by counting the number of subsequent processes that each process depends on and the total processing time. The initial priority value is obtained by multiplying the reciprocal of the remaining leeway time by the path urgency, and then by the material shortage risk coefficient of the process. The initial priority value is normalized, and the output is a dynamic priority index between 0 and 1.

4. The method for optimizing an automated production line based on artificial intelligence according to claim 3, characterized in that, The material shortage risk coefficient is dynamically calculated based on the historical supply volatility of the materials corresponding to the process and the arrival confidence of the current purchase orders in transit.

5. The method for optimizing an automated production line based on artificial intelligence according to claim 1, characterized in that, The identification of process blockage nodes caused by equipment conflicts includes: Iterate through the current process sequence and check for each process whether its matching equipment has been occupied by other processes within the planned execution time window; If the device is occupied, the process of the occupied matching device will be marked as a blocked node, and the conflicting device identifier and conflict time interval will be recorded. Calculate the waiting margin of the blocked node, which is the difference between the latest start time of the process that can be postponed on the occupied matching equipment and the current planned start time; When the waiting margin is less than the preset threshold, the blocked node and its subsequent dependent processes are marked as a set to be adjusted.

6. The method for optimizing an automated production line based on artificial intelligence according to claim 1, characterized in that, The method of adjusting the execution order of blocked processes using a backtracking conflict resolution strategy includes: From the processes already scheduled on the conflicting equipment, select the process with the lowest dynamic priority index as the removable process; Remove the removable process from the original equipment queue and place it in the reassignment pool; Re-execute the process equipment capacity matching step for the removable process, and assign another idle equipment to the removable process or postpone it to the next idle time slot of the equipment; If no feasible allocation can be found, backtrack to the previous conflict node and recursively perform the removal and reallocation operation until all conflicts are resolved.

7. The method for optimizing an automated production line based on artificial intelligence according to claim 1, characterized in that, The feasibility verification of the process sequence based on material kitting constraints includes: Create a bill of materials for each process, including the types, specifications, and quantities of raw materials required; Query the real-time inventory of the current material inventory table, and add the remaining quantity after all materials consumed before the target process; If the remaining quantity of any material required for the target process is less than the requirement, then the target process is determined not to meet the material completeness condition. Record the estimated replenishment time for the missing material and move the planned start time of the process to the earliest time slot after the replenishment time.

8. The method for optimizing an automated production line based on artificial intelligence according to claim 7, characterized in that, The method of postponing the planned start time of the target process includes: Obtain confirmation of delivery time and safety margin from suppliers for the materials in short supply, and calculate the latest available time. Based on the latest available time, find the first idle time slot of the candidate equipment for the target process after the latest available time in the process equipment capacity matching diagram; Adjust the planned start time of the target process to the start time of the first idle time slot, and synchronously update the start time of the subsequent dependent processes of the target process, using topological order to postpone them sequentially.

9. The method for optimizing an automated production line based on artificial intelligence according to claim 7, characterized in that, The material completeness condition also includes compatibility verification of substitute materials. When the main material is in short supply, if the substitute material meets the same specifications and has sufficient quantity, it is determined that the target process still meets the completeness condition.

10. The method for optimizing an automated production line based on artificial intelligence according to claim 8, characterized in that, If, during the sequential postponement of the topological order, the postponement of any subsequent dependent process causes the planned completion time to exceed the order delivery date, an emergency order insertion warning will be triggered and the backtracking conflict resolution strategy will be re-executed.

Images