Automobile front axle and rear axle processing production line intelligent scheduling method

Abstract

The application discloses an intelligent scheduling method for automobile front axle and rear axle processing production lines, and particularly relates to the technical field of manufacturing production scheduling, which comprises the following steps: obtaining a production performance flow with a workpiece traceability number throughout, recursively deriving an event position according to the chronological order of the recorded events, positioning a floating attribution key starting segment with the recorded events, positioning a floating attribution key attribution segment with the performance events, positioning a floating attribution key time sequence segment with the final processing events, and outputting a floating record to be scheduled; performing floating merging on the front axle traceability number and the rear axle traceability number of the same specification without shipping confirmation; recursively deriving the production performance events with the workpiece traceability number, constructing a floating pool of the same specification, and determining the target order segment writing depth by combining the particle Markov chain Monte Carlo algorithm, the Lagrange list generation algorithm and the reinforcement learning scheduling strategy, so that the front axle traceability number and the rear axle traceability number are kept in floating association before shipping confirmation and are written into the order performance batch after meeting the complete shipping conditions.

Description

Technical Field

[0001] This invention relates to the field of production scheduling technology in the manufacturing industry, and more specifically, to an intelligent scheduling method for automotive front and rear axle processing production lines. Background Technology

[0002] In the production lines for front and rear axles of automobiles, the existing scheduling is mostly based on order tracking, process connection and complete set shipment. The production management platform usually links specific front axle parts and specific rear axle parts with order numbers, vehicle models and shipment batch numbers when the blanks are put on the line, the first process is reported or barcode files are created. Then, the subsequent processing is arranged according to the delivery date, equipment availability, process sequence and complete set inspection results. In mixed production of multiple orders for the same model, although front axle and rear axle parts have the same specification replacement space, the re-testing of the hole system after rough boring of the rear axle housing, the repair of the mounting surface, the confirmation of fixture reset, and the final inspection return will constantly change the subsequent processing path. At the same time, the finished product buffer space is limited, the shipment batches cannot be separated, and the modification of traceability barcodes is controlled, making it difficult to readjust the fixed associations formed in the early stage according to the on-site processing results. As a result, it can be directly seen in the workshop production scheduling that when a certain order's rear axle is delayed due to retesting or repair, the qualified front axle bound to it can only occupy the buffer and wait. Urgent orders of the same specifications cannot be shipped in full due to the lack of front axles. Even if the planners repeatedly adjust the equipment sequence and test queue, they can only make up for it within the existing relationship and cannot release the qualified parts that have been locked too early. The technical problem this application aims to solve is: how to incorporate the timing of the complete set association between specific front axle components and specific rear axle components into intelligent scheduling decisions in the automotive front and rear axle processing production line, so as to avoid order lock-up caused by early fixed association, maintain the allocation leeway for processed parts of the same specification before the uncertain processing results are clear, and improve the stability of the complete set shipment of front and rear axles. Summary of the Invention

[0003] To overcome the aforementioned deficiencies in the prior art, embodiments of the present invention provide an intelligent scheduling method for automotive front and rear axle processing production lines. This method constructs a floating pool of the same specifications by recursively inferring production fulfillment events using workpiece traceability numbers. It also combines a particle Markov chain Monte Carlo algorithm, a Lagrange column generation algorithm, and a reinforcement learning scheduling strategy to determine the writing depth of the target order segment. This ensures that the front axle traceability number and the rear axle traceability number remain floatingly associated before shipment confirmation and are written into the order fulfillment batch after the complete set shipment conditions are met, thereby solving the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an intelligent scheduling method for automotive front and rear axle processing production lines, comprising: S1. Obtain the production fulfillment flow with the workpiece traceability number, and recursively push the event position according to the order of events from filing to the present. Use the filing event to locate the floating attribution key start segment, use the fulfillment event to locate the floating attribution key attribution segment, and use the final processing event to locate the floating attribution key time sequence segment. Output the floating record to be scheduled. S2. Perform floating merging on the front axle traceability number and rear axle traceability number of the same specification that have not been confirmed for shipment. Calculate from the current processing order to the shipment confirmation position along the process sequence, write the number of unfinished digits into the unassembled distance, and output the same specification floating pool. S3. Based on the floating pool of the same specification, execute the particle Markov chain Monte Carlo algorithm. Take the order segment writing depth as the first segment of the particle hidden state, take the review return event bit access repair occupied event bit as the last segment of the particle hidden state, and divide the number of the same position of the actual sequence and the predicted sequence by the total number of predicted sequences to generate the particle weight, and output a set of associated posterior records. S4. Based on the complete set of associated posterior records, execute the Lagrange column generation algorithm, write the order segment into the deep access manufacturing resource planning ledger and convert it into a resource scheduling column. Deduct the revenue from complete set shipment based on the number of repeated resource time slots and the number of single-end unmatched items, and output the column retrieval value. S5. Update the value of resource scheduling actions by using reinforcement learning scheduling strategy. The order segment write depth is the scheduling action, the complete set shipment increment is the action gain, and the single-end cache increment access to repair occupation increment is the action loss. The target order segment write depth is generated recursively.

[0005] In a preferred embodiment, it further includes: S6. Write the target order segment into the same specification floating pool for deep update. Perform associated segment migration on the paired traceability number according to the shipment confirmation position. Make the paired traceability number that has not crossed the shipment confirmation position stay in the withdrawn associated segment. Make the paired traceability number that has reached the shipment confirmation position migrate into the order fulfillment batch. Output the scheduling results of the automotive front and rear axle processing production line for enterprise or organization manufacturing resource fulfillment scheduling.

[0006] In a preferred embodiment, S1 includes: S1-1. In the production fulfillment flow, the first documented event is locked by the workpiece traceability number. The event bit of the first documented event is used as the starting point for recursion, and subsequent events are connected in ascending order according to the event time to form an event candidate chain. S1-2. For adjacent events in the event candidate chain, perform an equivalence comparison between the preceding event bit carried by the subsequent event and the event bit of the preceding event. If the equivalence comparison matches, add the subsequent event to the recursive chain. If the equivalence comparison does not match, write the subsequent event to the pending bit and output the final valid event bit. S1-3. Based on the last valid event bit backtracking recursion chain, write the floating ownership key start segment with the recursion start point, rewrite the floating ownership key ownership segment with the fulfillment ownership update bit, write the floating ownership key timing segment with the last processing order bit, and output the floating record to be scheduled.

[0007] In a preferred embodiment, S2 includes: S2-1. Read the starting segment of the floating ownership key in the floating record to be scheduled, connect the traceability number with the same starting segment and empty shipping confirmation bit to the candidate chain of the same specification, and write the traceability number into the corresponding bridge direction candidate chain according to the bridge direction mark, and output the bridge direction candidate chain. S2-2. For each traceability number in the bridge candidate chain, take the current processing order as the starting position, increment the starting position along the process number to the shipment confirmation position, and increment the bit register word by one every time an unreported process position is passed, and write the bit register word as the traceability distance. S2-3. Based on the traceability distance, perform ascending ranking of the bridge candidate chains, write the traceability numbers of the front bridge and the rear bridge with the same ranking into the same floating cell, write the traceability distance that lacks a matching traceability number with the same ranking into the unmatched distance, and output the floating pool of the same specification.

[0008] In a preferred embodiment, S3 includes: S3-1. In the same specification floating pool, locate the order segment with the traceability number of the incomplete set distance and write it into the depth candidate bit. Write the order segment into the depth candidate bit and write it into the first segment of the hidden state of the particle. Write the interval number obtained by recursively pushing the review return event bit to the rework occupied event bit along the event sequence into the tail segment of the hidden state of the particle. Output the initial state chain of the particle. S3-2. For the initial state chain of particles, the hidden state of the particle at the previous event position is taken as the ancestor state. The processing order change of the current event position is superimposed on the ancestor state to obtain the Markov proposal state. The proposal error word is generated according to the position difference between the predicted order and the actual processing order of the Markov proposal state, and the particle proposal chain is output.

[0009] In a preferred embodiment, S3 further includes: S3-3. Perform particle Markov chain Monte Carlo acceptance operation based on particle proposal chain. Connect the proposal error word to the single-end write count to generate the proposal cost word. Connect the ancestor state error word to the ancestor state single-end write count to generate the ancestor cost word. Use the hidden state corresponding to the cost word in lexicographical order as the hidden state of the next event bit particle and output the particle recursion chain. S3-4. Perform weight normalization and convergence writing according to the particle recursion chain. Compare the actual processing order with the predicted order bit by bit to obtain the same number of bits. Divide the same number of bits by the number of bits in the predicted order to obtain the particle weight. Divide the weight of each particle by the sum of the particle weights to obtain the normalized weight. When the hidden state number string of the particle in this round is the same as the hidden state number string of the particle in the previous round bit by bit, output a set of associated posterior records.

[0010] In a preferred embodiment, S4 includes: S4-1. Read the order segment write depth in the complete set of associated post-test records according to the normalized weight position order, extract the pre-shipment process segment with the paired traceability number in the same specification floating pool based on the order segment write depth, and write the pre-shipment process segment bit by bit into the resource time period cell in the manufacturing resource planning ledger, and output the candidate resource column. S4-2. Perform principal column verification on the candidate resource column using the Lagrange column generation algorithm. Subtract one from the number of times the traceability number is written in the resource time period cell to obtain the resource slack. Obtain the matching slack by subtracting the number of times the single-end traceability number is written in the shipment batch. Update the principal column multiplier with the resource slack and the matching slack.

[0011] In a preferred embodiment, S4 further includes: S4-3. Write back the candidate resource column using the main column multiplier, multiply the main column multiplier by the resource slack to obtain the resource deduction word, multiply the main column multiplier by the matching slack to obtain the matching deduction word, and deduct the resource deduction word and the matching deduction word in turn using the complete set shipment revenue, and output the candidate column search word. S4-4. Perform column filling based on the candidate column search word. Merge the candidate resource column with the first candidate column search word value into the resource scheduling column. Compare the merged resource scheduling column number with the resource scheduling column number of the previous round bit by bit. Stop column filling when all bit numbers are the same and output the column search value.

[0012] In a preferred embodiment, S5 includes: S5-1. Read the resource scheduling column based on the column retrieval value, write the order segment in the resource scheduling column into the scheduling action bit with the writing depth, and subtract the number of order fulfillment batches before the resource scheduling column is executed from the number of order fulfillment batches after the resource scheduling column is executed to obtain the complete set shipment increment. S5-2. Calculate the action reward of the scheduling action bit through the reinforcement learning scheduling strategy, write the complete set shipment increment into the benefit word, connect the single-end cache occupancy bit newly added after the resource scheduling column is executed to the repair occupancy bit to form the loss word, and use the benefit word to subtract the loss word to obtain the action reward of this round. S5-3. Use the original action value corresponding to the scheduling action bit to access the action reward of this round and the action value of the next event bit to generate an updated action value. Select the corresponding order segment write depth according to the bit order result of the updated action value and output the target order segment write depth.

[0013] In a preferred embodiment, S6 includes: S6-1. Read the matching traceability number of the same grid position in the floating pool of the same specification according to the target order segment write depth. Write the process bit chain of the matching traceability number from the current processing order to the target order segment write depth bit by bit. Increment the migration step number by one for each process bit written. When encountering an unreported process bit, write the unreported process bit to the process bit to be supplemented and stop the current round of writing. Output the associated migration word. S6-2. Calculate the shipping difference word based on the associated migration word. Subtract the shipping confirmation bit from the writing depth of the target order segment to obtain the shipping difference word. When the shipping difference word is negative, write the matching traceability number into the withdrawing associated segment. When the shipping difference word is not negative and the pending process bit is empty, write the matching traceability number into the order fulfillment batch and output the batch migration word. S6-3. Rewrite the manufacturing resource fulfillment scheduling table according to the batch migration word, fill the matching traceability number in the withdrawn associated segment back into the floating pool of the same specification, write the matching traceability number in the order fulfillment batch into the shipping schedule position, and generate the automotive front and rear axle processing production line scheduling result with the shipping schedule position after writing.

[0014] The technical effects and advantages of this invention are as follows: 1. This solution controls the timing of front and rear axle assembly association by writing the target order segment into the depth control, so that workpieces of the same specification remain in a floating and merged state before the review and rework results are clear, which relatively alleviates the order lock-up and single-end inventory occupation problems caused by early fixed association; 2. The production fulfillment flow is recursively generated into a floating record to be scheduled based on the event position, so that the workpiece traceability number, processing sequence and fulfillment source maintain the event continuity, and reduce the interference of supplementary events and chain break events on the floating merging results; 3. Calculate the distance of non-set for the traceability numbers of the front and rear axles of the same specification, and write it into the floating grid, so that the scheduling object changes from the order of completion of individual pieces to the depth of set writing, which improves the pairing selection process. 4. In the particle Markov chain Monte Carlo simulation, the verification return event bit and the rework occupancy event bit are added to make the set of related decisions combined with the uncertain changes in processing, which relatively reduces the risk of incorrect binding caused by rework lag. 5. The order segment write depth is transformed into a resource scheduling column using the Lagrange column generation algorithm, so that the revenue from complete shipment, resource occupation, and the situation of missing single-end components can all participate in the column retrieval, thereby improving the manufacturing resource allocation results. 6. Based on the reinforcement learning scheduling strategy, the incremental value of operation is pushed forward, so that the incremental value of complete set shipment, single-end cache, and repair occupation jointly determine the write depth of the target order segment, which relatively suppresses the scheduling deviation of only pursuing single-item advancement. Attached Figure Description

[0015] Figure 1 This is a flowchart outlining the method steps of the present invention. Detailed Implementation

[0016] 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.

[0017] Refer to the instruction manual appendix Figure 1 The intelligent scheduling method for automotive front and rear axle processing production lines of the present invention includes: S1. Obtain the production fulfillment flow with the workpiece traceability number, and recursively push the event position according to the order of events from filing to the present. Use the filing event to locate the floating attribution key start segment, use the fulfillment event to locate the floating attribution key attribution segment, and use the final processing event to locate the floating attribution key time sequence segment. Output the floating record to be scheduled. In this implementation, S1 is used to organize manufacturing production fulfillment events into floating records to be scheduled that S2 can read. This ensures that each workpiece traceability number first forms a single recursive path, and then the recursive path is written to the floating attribution key. This prevents supplementary events, duplicate reporting events, or attribution rewriting events from directly entering the same specification floating pool, thus avoiding incorrect merging of the preceding and following bridge traceability numbers. The production fulfillment log is generated by the production management platform during workpiece filing, process reporting, quality verification, rework registration, temporary storage, and shipment scheduling. Each production fulfillment event carries a workpiece traceability number, event bit, event time, preceding event bit, business attribution bit, and processing order. The event bit is generated according to the writing order of the production fulfillment log, and the preceding event bit is written from the previously confirmed event bit read when the current production fulfillment event is generated. This implementation process includes the following steps: S1-1 is used to establish a recursive starting point and candidate sorting path for each workpiece traceability number in the production fulfillment flow, so that subsequent event bit recursion has a single entry point. The production fulfillment flow is read using the workpiece traceability number as the search key. The earliest written production fulfillment event with the event category of filing is first located as the first filing event, and the event bits carried by the first filing event are written to the recursive starting point. Then, production fulfillment events under the same workpiece traceability number that are later than the recursive starting point are read and added to the event candidate chain in ascending order of event time. Events with the same event time are added according to the order in which event bits are written. For duplicate event bits, the one referenced by the preceding event bit of the subsequent production fulfillment event is retained. Other duplicate events are written to the pending position and do not enter the event candidate chain. If no first filing event is found for the same workpiece traceability number, no recursive starting point is generated. The workpiece traceability number is written to the missing filing record and the generation of floating records to be scheduled in this round is stopped. S1-2 is used to verify the actual connection relationship of the event candidate chain using the preceding event bit, so that the supplementary event and the broken chain event do not rewrite the final valid event bit. For adjacent production fulfillment events in the event candidate chain, the preceding event bit carried by the subsequent event is read and an equivalent comparison is performed with the event bit of the preceding event. If the equivalent comparison results are the same, the subsequent event is added to the recursive chain, and the event bit of the subsequent event is updated to the current valid event bit of the recursive chain. If the equivalent comparison results are different, the subsequent event along with the preceding event bit carried by the subsequent event is written to the suspended position, and the recursive chain still retains the previous current valid event bit. If the preceding event bit of the subsequent event is empty, the subsequent event is written to the suspended position, and it waits for the preceding event bit of the same workpiece traceability number to be added later before participating in the event candidate chain generation again. After the event candidate chain processing is completed, the event bit that was last completed in the recursive chain is written as the final valid event bit for S1-3 to backtrack and read. S1-3 is used to compress the recursive chain corresponding to the final valid event bit into a floating ownership key, so that S2 generates a floating pool of the same specification according to the manufacturing order fulfillment relationship and processing order; based on the final valid event bit, backtracking along the recursive chain to the recursive starting point, the vehicle model specification code in the first filing event corresponding to the recursive starting point is written into the floating ownership key starting segment; in the recursive chain, the fulfillment ownership update bit is read from the recursive starting point to the final valid event bit. When a new fulfillment ownership update bit is read, the floating ownership key ownership segment is overwritten. When no fulfillment ownership update bit is read, the initial order ownership bit in the first filing event is written into the floating ownership key ownership segment. The process reads the processing order in the production fulfillment event corresponding to the final valid event bit and writes the processing order into the floating home key time sequence segment. If the production fulfillment event corresponding to the final valid event bit does not carry a processing order, the process reads the most recent production fulfillment event carrying a processing order in reverse along the recursive chain and writes it into the floating home key time sequence segment. If the processing order is still not read, the workpiece traceability number is written into the order missing record. After completing the writing of the start segment, home segment, and time sequence segment, a floating record to be scheduled is generated, and the suspended position, missing file record, and missing order record are retained as abnormal auxiliary records and are not written into the same specification floating pool of S2. Through the above implementation process, the production fulfillment flow first goes through the event candidate chain sorting, then through the preceding event bit verification to form a recursive chain, and finally through the end valid event bit backtracking to form a floating attribution key, so that the vehicle specification source, order fulfillment source, and processing order in the floating record to be scheduled all have an event source; the suspended position only saves the broken chain or duplicate production fulfillment events, and does not participate in the formation of the floating pool of the same specification, thereby preventing the same workpiece traceability number from being incorrectly merged due to supplementary reporting, duplicate reporting, or abnormal attribution rewriting; In practical applications: After a certain rear axle traceability number is filed, it sequentially generates a rough processing report event, a quality review event, and a rework registration event. The preceding event bit carried by the rework registration event is the same as the event bit of the quality review event. Then, the rework registration event is connected to the recursive chain and becomes the final valid event bit. If at the same time, another temporary storage entry event is added, and the preceding event bit still points to the rough processing report event, then the temporary storage entry event is written to the pending bit, without changing the floating ownership key time sequence segment. The floating record to be scheduled read by S2 still uses the processing order corresponding to the rework registration event as the current processing order.

[0018] S2. Perform floating merging on the front axle traceability number and rear axle traceability number of the same specification that have not been confirmed for shipment. Calculate from the current processing order to the shipment confirmation position along the process sequence, write the number of unfinished digits into the unassembled distance, and output the same specification floating pool. In this embodiment, S2 is used to convert the floating records to be scheduled output by S1 into floating pools of the same specification. This allows the front axle traceability numbers and rear axle traceability numbers that have not yet entered shipment confirmation to first form a candidate range according to vehicle model specifications, and then form a traceability distance according to the remaining positions in the process sequence. The floating grid retains the allocation space before complete set association. The starting segment of the floating attribution key in the floating records to be scheduled serves as the source of same specification determination, the floating attribution key timing segment serves as the source of the current processing order, the shipment confirmation bit is given by the process sequence table of the corresponding vehicle model specification code, and the bridge direction mark is generated by the workpiece category bit in the first filing event. This implementation process includes the following steps: S2-1 is used to filter out traceability numbers of the same specification that have not yet been confirmed for shipment from the floating records to be scheduled, and generate the bridge direction candidate chain required for subsequent pairing according to the bridge direction; read the starting segment of the floating attribution key in the floating records to be scheduled, and write the traceability numbers with the same starting segment into the candidate chain of the same specification; then read the shipment confirmation bit of each traceability number. When the shipment confirmation bit is empty, it is kept in the candidate chain of the same specification. When the shipment confirmation bit has been written into the order fulfillment batch, it is removed from the candidate chain of the same specification; then read the workpiece category bit in the first filing event. When the workpiece category bit points to the front axle, the traceability number is written into the front axle candidate chain. When the workpiece category bit points to the rear axle, the traceability number is written into the rear axle candidate chain. When the workpiece category bit is empty or does not match the vehicle model specification code corresponding to the starting segment of the floating attribution key, the traceability number is written into the bridge direction abnormal record and does not enter the bridge direction candidate chain. Finally, the front axle candidate chain and the rear axle candidate chain are output for S2-2 to read. S2-2 is used to convert each traceability number in the bridge candidate chain into a traceability distance that can participate in the sorting, so that the traceability numbers of the front bridge and the back bridge with different process progresses can obtain the distance expression under the same process sequence table. For each traceability number in the front bridge candidate chain and the back bridge candidate chain, the process sequence table corresponding to the floating home key start segment is read, and the current processing order written in the floating home key timing segment is used as the starting position. The process number is incremented bit by bit along the process position number in the same process sequence table until the shipment confirmation position. The initial value of the bit register is written to zero. The bit register is incremented by one every time an unreported process position is passed. The unreported process position is formed by the absence of the corresponding process position reporting event in the production fulfillment flow, or the corresponding process position reporting event has entered the pending position. If the current processing order is not in the process sequence table, the traceability number is written to the order exception record and the traceability distance generation is stopped. After incrementing to the shipment confirmation position, the bit register is written as the traceability distance, and the traceability distance is written back to the corresponding bridge candidate chain along with the traceability number. S2-3 is used to write the front axle candidate chain and the rear axle candidate chain into the floating cell based on the traceability distance, so that the floating pool of the same specification stores both the traceability numbers that have been paired and the incomplete distance formed by the single-end missing pair. The front axle candidate chain and the rear axle candidate chain are ranked in ascending order by traceability distance. If the traceability distances are the same, they are ranked in ascending order by the end position valid event bit. If the end position valid event bits are the same, they are ranked in the traceability number encoding order. After the ranking is completed, the front axle traceability number and the rear axle traceability number in the same ranking are written into the same floating cell, and the absolute value of the difference between the two traceability distances is written into the incomplete distance of the floating cell. If there is only a front axle traceability number or only a rear axle traceability number in the same ranking, the single-end traceability number is written into the floating cell, and the traceability distance of the single-end traceability number is written into the incomplete distance. After all floating cells are written, the floating pool of the same specification is output, so that S3 can locate the order segment and write the depth candidate bit by the incomplete distance. Through the above implementation process, S2 transforms the same specification source, bridge direction source, and current processing order in the floating record to be scheduled into same specification candidate chain, bridge direction candidate chain, traceability distance, and floating grid, so that the front bridge traceability number and the rear bridge traceability number still retain space to be re-paired according to the processing progress before officially entering the order fulfillment batch; the traceability number whose shipment confirmation bit has been written into the order fulfillment batch will no longer enter the same specification floating pool, and the traceability number with missing bridge direction mark, missing process sequence table, or current processing order that cannot fall into the process sequence table will enter the abnormal record and will not participate in the complete set association post-verification calculation of S3; In practical applications: Under the same vehicle model specification code, there are three front axle traceability numbers and two rear axle traceability numbers. S2 first removes the traceability numbers that have been written into the order fulfillment batch, and then generates the front axle candidate chain and the rear axle candidate chain according to the bridge direction mark. One of the front axle traceability numbers has two unreported process positions between the current processing order and the shipment confirmation position, and the traceability distance is written as two. One of the rear axle traceability numbers has three unreported process positions, and the traceability distance is written as three. After the two are written into the same floating cell in the same position, the un-set distance is written as one. When the remaining front axle traceability number does not have a rear axle traceability number in the same position, the single-end front axle traceability number is written into another floating cell, and the corresponding traceability distance is written into the un-set distance.

[0019] S3. Based on the floating pool of the same specification, execute the particle Markov chain Monte Carlo algorithm. Take the order segment writing depth as the first segment of the particle hidden state, take the review return event bit access repair occupied event bit as the last segment of the particle hidden state, and divide the number of the same position of the actual sequence and the predicted sequence by the total number of predicted sequences to generate the particle weight, and output a set of associated posterior records. In this implementation, S3 is used to convert the incomplete set distance in the same specification floating pool into the posterior judgment result of the order segment write depth, so that the front bridge traceability number and the rear bridge traceability number undergo event bit inference, rework occupancy inference and sequence deviation comparison before entering the order fulfillment batch; the particle Markov chain Monte Carlo algorithm runs at the production fulfillment event level, does not change the equipment control parameters, and only processes the order segment write depth, processing sequence, review return event bit, rework occupancy event bit and single-end write count, so that the subsequent S4 can read the complete set associated posterior record and generate the resource scheduling column; the implementation process includes the following steps: S3-1 is used to convert floating cells in a floating pool of the same specification into a particle initial state chain, so that each order segment write depth candidate bit has an event source and a return flow occupancy source; it reads the unassembled distance of each floating cell in the floating pool of the same specification, and locates the order segment write depth candidate bit according to the displacement position corresponding to the unassembled distance in the process bit chain between the current processing sequence and the shipment confirmation position; if the unassembled distance is greater than the number of process bits between the current processing sequence and the shipment confirmation position, then the shipment confirmation position is written as the order segment write depth candidate bit; then the order segment write depth candidate bit is written into the particle hidden state first segment; and the return flow event is verified. The component position is written by the quality review event, and the rework occupation event position is written when the rework task occupies manufacturing resources. When there are review return event positions and rework occupation event positions, the event sequence of the production fulfillment flow is followed from the review return event position to the rework occupation event position. Each time an event position is passed, the interval number is incremented by one, and the interval number is written to the hidden state tail segment of the particle. When there are no review return event positions, the hidden state tail segment of the particle is written to zero. When there are review return event positions but no rework occupation event position is formed, the current event position is used as the recursion endpoint to generate the interval number. After completing the writing of the hidden state first segment and the hidden state tail segment of the particle, the initial state chain of the particle is output. S3-2 is used to modify the initial state chain of particles using the current production fulfillment event, so that the order segment write depth candidate bit follows the actual processing sequence change to form a Markov proposal state; for each particle in the initial state chain, the hidden state of the particle at the previous event bit is read as the ancestor state, and the processing sequence corresponding to the previous event bit and the processing sequence corresponding to the current event bit are read. The processing sequence corresponding to the current event bit is subtracted from the processing sequence corresponding to the previous event bit to obtain the processing sequence change; when the processing sequence change is positive, the order segment write depth candidate bit in the ancestor state is shifted by the same number of bits along the process sequence table towards the shipment confirmation bit. The process bit after the action is written as the Markov proposal state prediction sequence; when the change in processing sequence is negative, the negative change is written into the backflow occupied part of the hidden state tail of the particle, and the order segment in the ancestor state is written into the depth candidate bit and retained as the Markov proposal state prediction sequence; when the change in processing sequence is zero, the Markov proposal state prediction sequence uses the order segment in the ancestor state and is written into the depth candidate bit; read the actual processing sequence corresponding to the current event bit, subtract the Markov proposal state prediction sequence from the actual processing sequence in the same process sequence table and take the absolute value to obtain the proposal error word, and output the particle proposal chain; S3-3 is used to perform the acceptance operation in the particle proposal chain, so that the order bias and single-ended writing jointly determine the hidden state of the next event bit particle. For each Markov proposal state in the particle proposal chain, the proposal error word is read, and the number of times only the front bridge trace number or only the rear bridge trace number is written under the same order segment write depth is counted to obtain the single-ended writing count. The proposal error word is written to the front segment of the proposal cost word, and the single-ended writing count is written to the back segment of the proposal cost word to form the proposal cost word. The ancestor state is read according to the same rules, and the ancestor state error word and the ancestor state single-ended writing count are formed to form the ancestor cost word. When comparing the proposal cost word and the ancestor cost word, the front segment value is compared first, and the cost word with the smaller front segment value comes first. If the front segment values ​​are the same, the back segment values ​​are compared, and the cost word with the smaller back segment value comes first. If the front segment values ​​and the back segment values ​​are the same, the ancestor state is retained. The hidden state corresponding to the lexicographically first cost word is written as the hidden state of the next event bit particle. After all particles have completed the acceptance operation, the particle recursion chain is output. S3-4 is used to transform the particle recursion chain into normalized weights and output a set of associated posterior records, so that subsequent resource scheduling column readings have a definite particle position order. For each particle in the particle recursion chain, the predicted position is read position by position along the predicted sequence chain from the current event position to the shipment confirmation position, and compared position by position with the actual processing position in the production fulfillment flow. If the position numbers are the same, the corresponding position consistency number is incremented by one. The predicted position number is the total number of predicted positions participating in the comparison. When the predicted position number is zero, the corresponding position consistency number is written as one and the predicted position number is also written as one. The corresponding position consistency number is divided by the predicted position number to obtain the particle weight, and then each particle weight is divided by the total particle weights to obtain the particle weight. The result is used to obtain the normalized weight; when the sum of all particle weights is zero, the normalized weight is written to the same allocation value in the order of particle numbers; the particle hidden state number string is formed by concatenating the particle number, the order segment written into the depth candidate position, and the particle hidden state tail segment. After the normalized weight is updated, the particle hidden state number string for this round is generated in the order of particle numbers; when the particle hidden state number string for this round is the same as the particle hidden state number string for the previous round bit by bit, the order segment is written into the depth candidate position, the normalized weight, and the particle hidden state tail segment are written into the set of associated posterior records; when the particle hidden state number string for this round has different bit numbers from the particle hidden state number string for the previous round, the process continues to execute S3-2 to S3-4 in the next event position; Through the above implementation process, S3 first converts the unfinished distance in the same specification floating pool into order segment write depth candidate bits, then forms particle initial state chain, particle proposal chain and particle recursion chain through event bit recursion, and finally forms a complete set association post-test record by normalized weight; the verification of return event bits, rework occupied event bits, processing sequence change, proposal error word and single-end write count all come from production fulfillment flow or the same specification floating pool, without introducing equipment motion control data; In practical applications: If the front axle trace number and the rear axle trace number within the same floating cell differ by two incomplete process bits, S3 shifts the current processing sequence two process bits to obtain the order segment write depth candidate bit; after quality review, the rear axle trace number enters the rework task. Between the review reflow event bit and the rework occupied event bit, three event bits are passed, so the particle hidden state tail segment is written as three; the next event bit shows that the actual processing sequence of the rear axle is backed up by one process bit. The Markov proposal state retains the order segment write depth candidate bit and adds the reflow occupied part. Then, the particle hidden state of the next event bit is determined by comparing the proposal cost word with the ancestor cost word. Finally, the order segment with the higher normalized weight is written into the complete set association posterior record for S4 to read.

[0020] S4. Based on the complete set of associated posterior records, execute the Lagrange column generation algorithm, write the order segment into the deep access manufacturing resource planning ledger and convert it into a resource scheduling column. Deduct the revenue from complete set shipment based on the number of repeated resource time slots and the number of single-end unmatched items, and output the column retrieval value. In this implementation, S4 is used to transform the complete set of associated post-hoc records output by S3 into candidate resource columns at the manufacturing resource level. This ensures that the order segment write depth does not directly enter the shipping schedule, but instead first undergoes resource time slot occupancy, shipping support occupancy, and main column multiplier deduction calculations to form the column retrieval value. The manufacturing resource planning ledger is generated by the resource calendar in the production management platform. The resource calendar records the production management occupancy status of processing stations, inspection stations, buffer positions, and shipping positions using resource numbers and resource time slots. S4 only processes manufacturing resource planning data and order fulfillment data, and does not involve machine tool motion control data. This implementation process includes the following steps: S4-1 is used to incorporate the order segment write depth from the complete set association post-record into the manufacturing resource planning ledger, ensuring that each candidate complete set relationship first forms a resource occupancy path; it reads the complete set association post-record and arranges the order segment write depth from largest to smallest according to the normalized weight; for order segment write depths with the same normalized weight, they are arranged in ascending order according to the position number of the order segment write depth in the process sequence table; then, using the order segment write depth as the starting position and the shipment confirmation position as the ending position, it extracts the process position chain of the paired traceability number within the same specification floating pool that has not yet been reported as completed. The process position is used to obtain the pre-shipment process segment. For each process position in the pre-shipment process segment, the resource number and resource time slot cell corresponding to the process position are read from the manufacturing resource planning ledger. The paired traceability number is written into the resource time slot cell according to the process position. If the resource time slot cell is occupied by the same candidate resource column, the writing is still retained and a duplicate writing record is generated. After all pre-shipment process segments are written, the candidate resource column is output. If the order segment writing depth has reached the shipment confirmation position, the pre-shipment process segment is empty, and the candidate resource column is only written to the shipment resource time slot cell corresponding to the shipment confirmation position. S4-2 is used to perform principal column verification on the candidate resource column, transforming resource occupancy conflicts and single-end occupancy of shipments into a basis for principal column multiplier recursion. The resource time period cells in the candidate resource column are read using the Lagrange column generation algorithm, and the number of times the traceability number is written within each resource time period cell is counted. If the number of writes is zero or one, the resource slack is written to zero; if the number of writes exceeds one, the number of writes is subtracted by one to obtain the resource slack. Then, the shipment batches in the candidate resource column are read. If both the front bridge traceability number and the rear bridge traceability number are written simultaneously within a shipment batch, the corresponding slack is written to zero. If only one traceability number is written within a shipment batch... When writing the front axle traceability number or only the rear axle traceability number, the number of times the single-end traceability number is written is used to obtain the matching slack amount; when performing the master column verification for the first time, both the resource multiplier and the matching multiplier of the previous round are written to zero. The resource multiplier of the current round is obtained by adding the resource slack amount to the resource multiplier of the previous round, and the matching multiplier of the current round is obtained by adding the matching slack amount to the matching multiplier of the previous round. The resource multiplier and the matching multiplier of the current round are written to the master column multiplier record; if no shipment batch is written in the candidate resource column, the matching slack amount is generated according to the number of times the single-end traceability number in the candidate resource column that has not entered the shipment confirmation position is written; S4-3 is used to write back the main column multiplier to the candidate resource column, so that the candidate resource column forms a sortable candidate column retrieval word; read the main column multiplier record, multiply the current round resource multiplier by the resource slack to obtain the resource deduction word, and multiply the current round matching multiplier by the matching slack to obtain the matching deduction word; the complete set shipment revenue is generated by the number of batches that have simultaneously written the front axle traceability number and the rear axle traceability number in the candidate resource column and reached the shipment confirmation position. Each batch formed increases the complete set shipment revenue by one. The single-end traceability number is not included in the complete set shipment revenue; then, the complete set shipment revenue is used to first deduct the resource deduction word, and then deduct the matching deduction word to obtain the candidate column retrieval word. The candidate column retrieval word is written into the candidate column retrieval table along with the candidate resource column number for S4-4 to fill in the execution column; if the candidate column retrieval word is negative, the candidate resource column is still retained in the candidate column retrieval table for subsequent column filling and comparison, and is not directly written into the resource scheduling column; S4-4 is used to perform column filling based on the candidate column search words, so that the candidate resource columns converge into resource scheduling columns and output the column search value; read the candidate column search table, arrange the candidate resource columns in descending order of the candidate column search word value, and arrange the candidate resource columns in ascending order of the candidate resource column number if the values ​​are the same; merge the first candidate resource column into the resource scheduling column, and concatenate the merging order with the candidate resource column number to generate the resource scheduling column number for this round; if the resource scheduling column number for this round has different position numbers from the resource scheduling column number for the previous round, then the resource scheduling column for this round is used as the main column basis for the next round of column filling, and S4-2 to S4-4 are executed again; if the resource scheduling column number for this round has all position numbers the same as the resource scheduling column number for the previous round, then column filling stops, the candidate column search word corresponding to the first candidate resource column is written as the column search value, and the resource scheduling column and the column search value are output to S5; during the first round of column filling, the resource scheduling column number for the previous round is written as an empty string, the resource scheduling column number is not stopped, and the next round of column filling is directly started; Through the above implementation process, S4 first transforms the order segment write depth in the complete set associated post-hoc record into a candidate resource column, then generates a principal column multiplier from the resource slack and the matching slack, and then forms a candidate column retrieval word through the resource deduction word, the matching deduction word, and the complete set shipment revenue. Finally, the resource scheduling column and column retrieval value are obtained through column supplementation. The generation of candidate resource columns, the recursion of principal column multipliers, the sorting of candidate column retrieval words, and the digit-by-digit comparison of resource scheduling column numbers all use the manufacturing resource planning ledger and the floating pool of the same specification as data sources. In practical applications: The write depth of a certain order segment corresponds to a set of front axle traceability numbers and rear axle traceability numbers. S4 first extracts the pre-shipment process segments that have not yet been reported and completed between the order segment write depth and the shipment confirmation position, and writes the pre-shipment process segments into the resource time slots corresponding to the processing station and the inspection station. If the same inspection resource time slot is written by two traceability numbers, the resource slack is written as 1. If only the front axle traceability number is written in the same shipment batch, the matching slack is written as 1. The main column multiplier generates the resource deduction word and the matching deduction word accordingly. After the candidate column retrieval word is deducted, the participating columns are added. Finally, the candidate resource columns that are arranged at the top and whose resource scheduling column numbers no longer change form the column retrieval value and are read by S5.

[0021] S5. Update the value of resource scheduling column actions by reinforcement learning scheduling strategy, taking the order segment write depth as the scheduling action, the complete set shipment increment as the action benefit, and the single-end cache increment access to repair occupation increment as the action loss, and recursively generate the target order segment write depth. In this implementation, S5 is used to convert the resource scheduling column output by S4 into the target order segment write depth, so that the reinforcement learning scheduling strategy does not directly output the equipment queuing result, but instead performs action value recursion based on the write depth of the front axle trace number and the rear axle trace number entering the order fulfillment batch; the resource scheduling column first performs a trial write in the manufacturing resource fulfillment scheduling table, and the trial write result is only used to calculate the action benefit and action loss, and does not rewrite the formal shipment schedule position before the target order segment write depth is output; this implementation process includes the following steps: S5-1 is used to convert the resource scheduling column corresponding to the column retrieval value into a scheduling action bit that the reinforcement learning scheduling strategy can read, and obtain the complete fulfillment change before and after the action execution; according to the column retrieval value output by S4, the resource scheduling column with the first position in the sorted list is read, the order segment write depth is read from the resource scheduling column, and the order segment write depth is written as a scheduling action bit; then, the current order fulfillment batch status is copied into the manufacturing resource fulfillment scheduling table as a trial write base table, the pairing traceability number in the resource scheduling column is written into the trial write base table according to the order segment write depth, the number of order fulfillment batches that simultaneously contain the front bridge traceability number and the rear bridge traceability number and have reached the shipment confirmation position before the trial write is counted, and the number of order fulfillment batches that simultaneously contain the front bridge traceability number and the rear bridge traceability number and have reached the shipment confirmation position after the trial write is counted, and the latter is subtracted from the former to obtain the complete shipment increment; if the order segment write depth is missing in the resource scheduling column, the resource scheduling column does not form a scheduling action bit, and the next resource scheduling column in the candidate column retrieval table is retrieved; S5-2 is used to compress changes in complete set fulfillment, single-end cache, and repair occupancy into the current round's action reward, enabling the reinforcement learning scheduling strategy to simultaneously penalize premature single-end writes and repair occupancy propagation. The reinforcement learning scheduling strategy reads the scheduling action bit and the trial write base table, writing the complete set shipment increment as an integer into the benefit word. It then reads the number of trace numbers that entered the cache resource time slot after the trial write but did not simultaneously form a front-bridge trace number and a rear-bridge trace number, writing this number into the single-end cache occupancy bit. It reads the number of trace numbers that still fell into the repair occupancy event bit after the trial write, writing this number into the repair occupancy bit. The single-end cache occupancy bit value is added to the repair occupancy bit value to obtain the loss word, and the benefit word value is subtracted from the loss word value to obtain the current round's action reward. If no cache resource time slot write is generated in the trial write base table, the single-end cache occupancy bit is written to zero; if no repair occupancy event bit is read, the repair occupancy bit is written to zero. S5-3 is used to perform action value recursion on the scheduled action bits and select the target order segment write depth from the recursion results for S6 to perform associated segment migration; read the original action value of the scheduled action bit in the action value table, and read the action value of the next event bit in the historical report table; if the next event bit has not yet generated a historical report, the action value of the next event bit is written to zero; add the original action value to the action report of this round and the action value of the next event bit to obtain the updated action value, and write the updated action value back to the action value table corresponding to the scheduled action bit; sort the updated action values ​​of all scheduled action bits in this round in descending order of value; when the updated action values ​​are the same, select the order segment write depth whose tag number is close to the shipment confirmation bit in the process sequence table; if the tag numbers are still the same, select according to the order segment write depth coding order; finally, output the selected order segment write depth as the target order segment write depth. Through the above implementation process, S5 first converts the resource scheduling column corresponding to the column retrieval value into a scheduling action bit, then obtains the complete set shipment increment, single-end cache occupancy bit and rework occupancy bit by writing through the manufacturing resource fulfillment scheduling table, and finally generates the target order segment write depth by recursively using the action value; this implementation process limits the action object of the reinforcement learning scheduling strategy to the order segment write depth, and the action reward to the changes in manufacturing order fulfillment and resource occupancy. In practical applications: After writing to the order segment write depth corresponding to a certain resource scheduling column, an order fulfillment batch simultaneously reaches the front bridge trace number and the back bridge trace number, and the complete set shipment incremental write is one; if the same trial write also causes a front bridge trace number to enter the cache resource time slot separately, and causes a back bridge trace number to continue to occupy the rework event bit, then the single-end cache occupied bit is written one, the rework occupied bit is written one, the loss word is written two, and the reward for this round of actions is negative one; accordingly, the action value table lowers the ranking of the order segment write depth, and instead selects another order segment write depth that can form a complete set shipment without adding a single-end cache as the target order segment write depth.

[0022] S6. Write the target order segment into the same specification floating pool for deep update, perform associated segment migration on the paired traceability number according to the shipment confirmation position, so that the paired traceability number that has not crossed the shipment confirmation position stays in the withdrawn associated segment, and the paired traceability number that has reached the shipment confirmation position is migrated into the order fulfillment batch, and output the scheduling results of the automotive front and rear axle processing production line for enterprise or organization manufacturing resource fulfillment scheduling; In this implementation, S6 is used to write the target order segment output by S5 into the same specification floating pool and the manufacturing resource fulfillment scheduling table, so that the paired traceability number remains in the withdrawn associated segment when the shipment confirmation condition is not met, and is migrated into the order fulfillment batch when the shipment confirmation condition is met; the paired traceability number in the same cell read by S6 comes from the same specification floating pool output by S2. A paired traceability number is formed when both the front bridge traceability number and the rear bridge traceability number exist in the same floating cell; the floating cell with only a single-end traceability number does not perform the order fulfillment batch migration and remains in the same specification floating pool; this implementation process includes the following steps: S6-1 is used to calculate the process bits traversed when the paired traceability number migrates from the current processing order to the target order segment write depth, and to truncate the current migration if there are unreported process bits; it reads the target order segment write depth and the paired traceability number in the same floating cell of the same specification floating pool, and then reads the process bit chain under the vehicle specification code corresponding to the paired traceability number; if the target order segment write depth is after the current processing order, it writes bit by bit from the current processing order along the direction of the shipment confirmation bit to the target order segment write depth, and the initial value of the migration step number is written to zero. The migration step number is incremented by one for each process bit written; during the writing process, the production fulfillment flow is read. If no reporting event occurs in the process bit or the reporting event enters the suspended position, the process bit is written to the process bit to be supplemented and the current writing is stopped; if the target order segment write depth is at or before the current processing order, the migration step number is written to zero and the paired traceability number maintains its original associated position; after the writing is completed, the paired traceability number, migration step number, process bit to be supplemented and target order segment write depth are concatenated to generate an associated migration word for S6-2 to read; S6-2 is used to determine whether the paired traceability number remains in the withdrawn associated segment or the migrated order fulfillment batch based on the associated migration word, and generates a batch migration word; it reads the target order segment write depth and the pending process bit in the associated migration word, and then reads the shipment confirmation bit in the process bit chain. According to the rule of increasing process bit number along the process sequence, the target order segment write depth is subtracted from the shipment confirmation bit to obtain the shipment difference word; when the shipment difference word is negative, it means that the target order segment write depth has not yet reached the shipment confirmation bit, and the paired traceability number, the target order segment write depth, and the pending process bit are written to the withdrawn associated segment; when the shipment difference word is not negative and the pending process bit is empty, and there are both front bridge traceability numbers and rear bridge traceability numbers in the same floating cell, the paired traceability number is written to the order fulfillment batch; when the shipment difference word is not negative but the pending process bit is not empty, the paired traceability number is still written to the withdrawn associated segment; after completing the judgment, the withdrawn associated segment write result or the order fulfillment batch write result is used to generate a batch migration word for S6-3 to rewrite the manufacturing resource fulfillment scheduling table; S6-3 is used to write the batch migration word into the manufacturing resource fulfillment scheduling table, and generate the scheduling result of the automotive front and rear axle processing production line from the shipment scheduling position. The manufacturing resource fulfillment scheduling table reads the resource scheduling column formed by S4, the target order segment write depth formed by S5, and the batch migration word formed by S6-2, and is used to receive the formal scheduling output after the target order segment write depth. When the batch migration word points to the withdrawn associated segment, the paired traceability number, target order segment write depth, and pending process position in the withdrawn associated segment are backfilled into the same specification floating pool, so that the paired traceability number continues to participate in floating merging in the next round of scheduling. When the batch migration word points to the order fulfillment batch, the paired traceability number in the order fulfillment batch is written into the shipment scheduling position, and the resource occupation record corresponding to the shipment scheduling position is written into the manufacturing resource fulfillment scheduling table. If the shipment scheduling position has already been written by the same order fulfillment batch, only the paired traceability number migration record is appended, and the shipment scheduling position is not generated again. After the writing is completed, the scheduling result of the automotive front and rear axle processing production line is generated with the shipment scheduling position, the order fulfillment batch, and the resource occupation record in the manufacturing resource fulfillment scheduling table. Through the above implementation process, S6 converts the target order segment write depth into associated migration words, shipping difference words, and batch migration words, and then completes the withdrawal of associated segment backfilling or order fulfillment batch writing through the manufacturing resource fulfillment scheduling table; when the pairing traceability number has not reached the shipping confirmation position, there is a pending process position, or there is a lack of bridge pairing in the same floating cell, it will not enter the order fulfillment batch; when the pairing traceability number reaches the shipping confirmation position and there is no pending process position, it will enter the shipping scheduling position and form the automotive front and rear axle processing production line scheduling result; In practical applications: The front axle traceability number and the rear axle traceability number have been paired in the same floating cell. The target order segment write depth output by S5 is two positions before the shipment confirmation position. After writing two process positions along the process position chain in S6-1, it is found that the process position corresponding to the rear axle traceability number has not been reported. Therefore, the process position of the rear axle that has not been reported is written to the process position to be supplemented. In S6-2, the shipment difference word is calculated to be negative. The paired traceability number is written to the withdrawn associated segment and backfilled into the floating pool of the same specification. In the next round of scheduling, the process position of the rear axle that has not been reported is reported. The target order segment write depth reaches the shipment confirmation position and the process position to be supplemented is empty. The paired traceability number is moved into the order fulfillment batch, and the corresponding automotive front and rear axle processing production line scheduling result is generated by the shipment scheduling position.

[0023] Furthermore, the present invention also includes, taking the same model specification "CX4600" of a certain automobile front and rear axle processing enterprise as an example, the process sequence table is arranged from G1 to G7, where G1 is the filing position, G4 is the quality review position, G5 is the rework position, and G7 is the shipment confirmation position; the production management platform has front axle traceability numbers F-A001 and F-A002 and rear axle traceability numbers R-A001 and R-A002. None of the four traceability numbers have been written into the shipment confirmation position. It is necessary to determine which set of front axle traceability numbers and rear axle traceability numbers should enter the order fulfillment batch, and which set is still in the same specification floating pool; In S1, the production management platform reads the production fulfillment log of traceability number F-A001. The first filing event is E100, and subsequent events E108, E116, and E124 point to the previous event position in sequence. After comparing the preceding event positions, all event candidate chains are connected to the recursive chain, and the final valid event position is E124, which corresponds to the processing sequence G5. In the production fulfillment log of traceability number R-A001, after filing event E101, work reporting event E109, quality review event E118, and rework registration event E127 are formed. Among them, the rework registration event E127 carries the preceding event... Since the part location is the same as the quality review event E118, E127 is connected to the recursive chain, and the final valid event bit is E127. The processing sequence corresponding to E127 is G4. The traceability numbers F-A002 and R-A002 also form the final valid event bits in the same way, and obtain the processing sequences G4 and G6 respectively. The production management platform then writes the vehicle specification code "CX4600" in the filing event into the floating attribution key start segment, writes the order fulfillment source into the floating attribution key attribution segment, writes the processing sequence corresponding to the final valid event bit into the floating attribution key timing segment, and outputs four floating records to be scheduled. In S2, the starting segment of the floating ownership key for all four floating records to be scheduled is "CX4600", and the shipment confirmation bit is empty. Therefore, the four traceability numbers are connected to the candidate chain of the same specification. The workpiece category bit in the first filing event writes F-A001 and F-A002 into the front bridge candidate chain and R-A001 and R-A002 into the rear bridge candidate chain. The traceability distance is calculated with G7 as the shipment confirmation bit. The current processing order of F-A001 is G5. G6 and G7 have not yet been reported for work, so the traceability distance is written as two. The current processing order of F-A002 is... For G4, G5, G6, and G7, which have not yet been reported for work, write three for the traceability distance; for R-A001, the current processing order is G4, and G5, G6, and G7 have not yet been reported for work, write three for the traceability distance; for R-A002, the current processing order is G6, and G7 has not yet been reported for work, write one for the traceability distance; after the front axle candidate chain and the rear axle candidate chain are sorted in ascending order of traceability distance, F-A001 and R-A002 are written into the first floating cell, and the distance of not being a complete set is written into one; F-A002 and R-A001 are written into the second floating cell, and the distance of not being a complete set is written into zero, and the same specification floating pool is output; In S3, the first floating cell in the same specification floating pool has a distance of one due to not being fully assembled, so the current processing order is shifted one process position to the next, and the order segment is written into the depth candidate position; the second floating cell has a distance of zero due to not being fully assembled, so the position corresponding to the current processing order is written into the depth candidate position as the order segment; R-A001 has a quality review event E118 and a rework registration event E127, and the production fulfillment flow is pushed from E118 to E127 through two event positions, so the particle's hidden tail is written as two; F-A002 has not had a rework registration, so the particle's hidden tail is written as zero; then, the hidden state of the particle at the previous event position is used as the ancestor state, and the change in processing order at the current event position is read to form a Markov reference. The proposal states; for example, R-A001's processing order did not advance to G7 due to repair registration. The Markov proposes that there is a position number difference between the predicted order and the actual processing order, and the proposal error word is written as 1. In the particle acceptance operation, the proposal error word of the second floating cell is 1, the number of single-end writes is zero, and the proposal cost word is "1-0". The first floating cell has one single-end write because R-A002 is close to the shipment confirmation position but F-A001 is still missing G6 reporting. The number of single-end writes is one, and the proposal cost word is "1-1". According to the comparison of the cost word with the first part priority and the second part continuation, the particle corresponding to the second floating cell enters the particle recursion chain, and after the weight is normalized, a set of associated posterior records with the normalized weight first is formed. In S4, the manufacturing resource planning ledger reads the order segment write depth corresponding to the second floating cell, extracts the pre-shipment process segments of F-A002 and R-A001 from the order segment write depth to G7 that have not yet been reported as completed, and writes them into the resource time period cell bit by bit; F-A002 needs to be written into the G5 finishing resource time period cell, the G6 temporary resource time period cell, and the G7 shipping resource time period cell; R-A001 needs to be written into the G5 rework resource time period cell, the G6 temporary resource time period cell, and the G7 shipping resource time period cell; if the G6 temporary resource time period cell is also written into F-A002 For R-A001, the resource slack is written as 1; since F-A002 and R-A001 enter the same shipment batch at the same time, the matching slack is written as zero; after the main column multiplier is recursively calculated using the resource slack and matching slack, the resource deduction word is 1, and the matching deduction word is zero; the complete shipment revenue is written as 1 because it forms a complete shipment batch, and the candidate column search word is zero; since the first floating cell will only form a single-end shipment of R-A002 first, the matching slack is written as 1, and the candidate column search word is lower than that of the second floating cell, so the candidate resource column corresponding to the second floating cell is merged into the resource scheduling column, and the output column search value is output. In S5, the reinforcement learning scheduling strategy reads the resource scheduling column corresponding to the retrieved column value, writes the order segment write depth corresponding to the second floating cell as the scheduling action bit, and performs a trial write in the manufacturing resource fulfillment scheduling table. Before the trial write, the order fulfillment batch number is zero. After the trial write, F-A002 and R-A001 arrive at the shipment confirmation bit at the same time and are written to the same order fulfillment batch, so the order fulfillment batch number is one, and the complete set shipment increment is written as one. After the trial write, neither F-A002 nor R-A001 forms a single-end cache occupation, and the single-end cache occupation bit is written as zero. Although R-A001 has a rework registration, the rework occupation event bit has been arranged in the resource scheduling column for the rework resource time slot, so the new quantity of rework occupation is written as zero, and the loss word is written as zero. The action return for this round is one, and the action value table places the order segment write depth corresponding to the second floating cell at the first position of the scheduling action bit for this round, and outputs the target order segment write depth. In S6, the manufacturing resource fulfillment scheduling table reads the target order segment write depth and the paired traceability numbers F-A002 and R-A001 in the second floating cell, and writes them bit by bit along the process bit chain from the current processing order to the target order segment write depth. If both G5 and G6 of F-A002 have obtained resource time slots in the resource scheduling column, the migration step number increases bit by bit. The G5 rework resource time slot of R-A001 has also been written to the resource scheduling column, and the process bit to be supplemented is empty. Then the shipping difference word is calculated. When the target order segment write depth reaches G7, the shipping difference word is not negative, and simultaneously within the same floating cell... There are front axle traceability numbers and rear axle traceability numbers. F-A002 and R-A001 are jointly written into the order fulfillment batch. F-A001 and R-A002 in the first floating cell are still backfilled into the same specification floating pool and retained in the withdrawal association segment because F-A001 still has an unreported process position. Finally, the manufacturing resource fulfillment scheduling table writes F-A002 and R-A001 into the shipping schedule position, generating the scheduling result of the automotive front and rear axle processing production lines. This round of scheduling did not lock F-A001 and R-A002 in advance, so that F-A001 still retains floating adjustment space before the subsequent G6 reporting is completed.

[0024] Working principle: This scheme first uses the workpiece traceability number to connect the production fulfillment flow of the front and rear axles. It identifies the current actual processing position of each workpiece through event bit recursion and generates a floating record to be scheduled. Then, the traceability numbers of the front and rear axles of the same specification that have not been confirmed for shipment are placed into a floating pool of the same specification, and the distance of non-setup is calculated according to the remaining process bits. Subsequently, the particle Markov chain Monte Carlo algorithm is used to infer the setup association results under different order segment write depths. Then, the Lagrange column generation algorithm is used to convert the candidate write depth into a manufacturing resource occupation scheme. Finally, the reinforcement learning scheduling strategy generates the target order segment write depth based on the set shipment increment, single-end buffer, and rework occupation, so that the front and rear axles remain floating before the conditions are ripe, and the order fulfillment batch is written after the shipment confirmation conditions are met. In practical applications, multiple front and rear axles of the same vehicle model are processed simultaneously on the production line. Some front axles are close to shipment, while some rear axles are still under review or repair. The solution does not bind specific front and rear axles when the blanks are put on the line or when they are reported for work in the early stages. Instead, it first determines the remaining process positions of each axle based on the production fulfillment events, and then prioritizes pairing the traceability numbers of front and rear axles that are close in progress, have appropriate resource usage, and can be shipped as a complete set. If a rear axle still has unreported process positions or is occupied by repairs, the corresponding pairing relationship remains in the withdrawal association segment and returns to the floating pool of the same specification. It will participate in scheduling again after the subsequent process is completed, thereby reducing the inventory of finished products at one end and the lock-in of orders.

[0025] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent scheduling method for automotive front and rear axle processing production lines, characterized in that, include: S1. Obtain the production fulfillment flow with the workpiece traceability number, and recursively push the event position according to the event sequence from the filing to the current. Use the filing event to locate the floating attribution key start segment, use the fulfillment event to locate the floating attribution key attribution segment, and use the final processing event to locate the floating attribution key time sequence segment. Output the floating record to be scheduled. S2. Perform floating merging on the front axle traceability number and rear axle traceability number of the same specification that have not been confirmed for shipment. Calculate from the current processing order to the shipment confirmation position along the process sequence, write the number of unfinished digits into the unassembled distance, and output the same specification floating pool. S3. Based on the floating pool of the same specification, execute the particle Markov chain Monte Carlo algorithm. Take the order segment writing depth as the first segment of the particle hidden state, take the review return event bit access repair occupied event bit as the last segment of the particle hidden state, and divide the number of the same position of the actual sequence and the predicted sequence by the total number of predicted sequences to generate the particle weight, and output a set of associated posterior records. S4. Based on the complete set of associated posterior records, execute the Lagrange column generation algorithm, write the order segment into the deep access manufacturing resource planning ledger and convert it into a resource scheduling column. Deduct the revenue from complete set shipment based on the number of repeated resource time slots and the number of single-end unmatched items, and output the column retrieval value. S5. Update the value of resource scheduling column actions by reinforcement learning scheduling strategy, taking the order segment write depth as the scheduling action, the complete set shipment increment as the action benefit, and the single-end cache increment access to repair occupation increment as the action loss, and recursively generate the target order segment write depth. S6. Write the target order segment into the same specification floating pool for deep update. Perform associated segment migration on the paired traceability number according to the shipment confirmation position. Make the paired traceability number that has not crossed the shipment confirmation position stay in the withdrawn associated segment. Make the paired traceability number that has reached the shipment confirmation position migrate into the order fulfillment batch. Output the scheduling result of the automotive front and rear axle processing production line for enterprise or organization manufacturing resource fulfillment scheduling. S2 includes: S2-1. Read the starting segment of the floating ownership key in the floating record to be scheduled, connect the traceability number with the same starting segment and empty shipping confirmation bit to the candidate chain of the same specification, and write the traceability number into the corresponding bridge direction candidate chain according to the bridge direction mark, and output the bridge direction candidate chain. S2-2. For each traceability number in the bridge candidate chain, take the current processing order as the starting position, and increment the starting position along the process number to the shipment confirmation position. Each time an unreported process position is passed, the bit register word is incremented by one, and the bit register word is written as the traceability distance. S2-3. Based on the traceability distance, perform ascending ranking of the bridge candidate chains, write the traceability numbers of the front and rear bridges with the same ranking into the same floating cell, write the traceability distances that lack matching traceability numbers with the same ranking into the unmatched distance, and output the floating pool of the same specification. S6 includes: S6-1. Read the matching traceability number of the same grid position in the floating pool of the same specification according to the target order segment write depth. Write the process bit chain of the matching traceability number from the current processing order to the target order segment write depth bit by bit. Increment the migration step number by one for each process bit written. When encountering an unreported process bit, write the unreported process bit to the process bit to be supplemented and stop the current round of writing. Output the associated migration word. S6-2. Calculate the shipping difference word based on the associated migration word. Subtract the shipping confirmation bit from the writing depth of the target order segment to obtain the shipping difference word. When the shipping difference word is negative, write the matching traceability number into the withdrawing associated segment. When the shipping difference word is not negative and the pending process bit is empty, write the matching traceability number into the order fulfillment batch and output the batch migration word. S6-3. Rewrite the manufacturing resource fulfillment scheduling table according to the batch migration word, fill the matching traceability number in the withdrawn associated segment back into the floating pool of the same specification, write the matching traceability number in the order fulfillment batch into the shipping schedule position, and generate the automotive front and rear axle processing production line scheduling result with the shipping schedule position after writing.

2. The intelligent scheduling method for automotive front and rear axle processing production lines according to claim 1, characterized in that: S1 includes: S1-1. In the production fulfillment flow, the first documented event is locked by the workpiece traceability number. The event bit of the first documented event is used as the starting point for recursion, and subsequent events are connected in ascending order according to the event time to form an event candidate chain. S1-2. For adjacent events in the event candidate chain, perform an equivalence comparison between the preceding event bit carried by the subsequent event and the event bit of the preceding event. If the equivalence comparison matches, add the subsequent event to the recursive chain. If the equivalence comparison does not match, write the subsequent event to the pending bit and output the final valid event bit. S1-3. Based on the last valid event bit backtracking recursion chain, write the floating ownership key start segment with the recursion start point, rewrite the floating ownership key ownership segment with the fulfillment ownership update bit, write the floating ownership key timing segment with the last processing order bit, and output the floating record to be scheduled.

3. The intelligent scheduling method for automotive front and rear axle processing production lines according to claim 2, characterized in that: S3 includes: S3-1. In the same specification floating pool, locate the order segment with the traceability number of the incomplete set distance and write it into the depth candidate bit. Write the order segment into the depth candidate bit and write it into the first segment of the hidden state of the particle. Write the interval number obtained by recursively pushing the review return event bit to the rework occupied event bit along the event sequence into the tail segment of the hidden state of the particle and output the initial state chain of the particle. S3-2. For the initial state chain of particles, the hidden state of the particle at the previous event position is taken as the ancestor state. The processing order change of the current event position is superimposed on the ancestor state to obtain the Markov proposal state. The proposal error word is generated according to the position difference between the predicted order and the actual processing order of the Markov proposal state, and the particle proposal chain is output.

4. The intelligent scheduling method for automotive front and rear axle processing production lines according to claim 3, characterized in that: S3 also includes: S3-3. Perform particle Markov chain Monte Carlo acceptance operation based on particle proposal chain. Connect the proposal error word to the single-end write count to generate the proposal cost word. Connect the ancestor state error word to the ancestor state single-end write count to generate the ancestor cost word. Use the hidden state corresponding to the cost word in lexicographical order as the hidden state of the next event bit particle and output the particle recursion chain. S3-4. Perform weight normalization and convergence writing according to the particle recursion chain. Compare the actual processing order with the predicted order bit by bit to obtain the same number of bits. Divide the same number of bits by the number of bits in the predicted order to obtain the particle weight. Divide the weight of each particle by the sum of the particle weights to obtain the normalized weight. When the hidden state number string of the particle in this round is the same as the hidden state number string of the particle in the previous round bit by bit, output a set of associated posterior records.

5. The intelligent scheduling method for automotive front and rear axle processing production lines according to claim 4, characterized in that: S4 includes: S4-1. Read the order segment write depth in the complete set of associated post-test records according to the normalized weight position order, extract the pre-shipment process segment with the paired traceability number in the same specification floating pool based on the order segment write depth, and write the pre-shipment process segment bit by bit into the resource time period cell in the manufacturing resource planning ledger, and output the candidate resource column. S4-2. Perform principal column verification on the candidate resource column using the Lagrange column generation algorithm. Subtract one from the number of times the traceability number is written in the resource time period cell to obtain the resource slack. Obtain the matching slack by subtracting the number of times the single-end traceability number is written in the shipment batch. Update the principal column multiplier with the resource slack and the matching slack.

6. The intelligent scheduling method for automotive front and rear axle processing production lines according to claim 5, characterized in that: S4 also includes: S4-3. Write back the candidate resource column using the main column multiplier, multiply the main column multiplier by the resource slack to obtain the resource deduction word, multiply the main column multiplier by the matching slack to obtain the matching deduction word, and deduct the resource deduction word and the matching deduction word in turn using the complete set shipment revenue, and output the candidate column search word. S4-4. Perform column filling based on the candidate column search word. Merge the candidate resource column with the first candidate column search word value into the resource scheduling column. Compare the merged resource scheduling column number with the resource scheduling column number of the previous round bit by bit. Stop column filling when all bit numbers are the same and output the column search value.

7. The intelligent scheduling method for automotive front and rear axle processing production lines according to claim 6, characterized in that: S5 includes: S5-1. Read the resource scheduling column based on the column retrieval value, write the order segment in the resource scheduling column into the scheduling action bit with the writing depth, and subtract the number of order fulfillment batches before the resource scheduling column is executed from the number of order fulfillment batches after the resource scheduling column is executed to obtain the complete set shipment increment. S5-2. Calculate the action reward of the scheduling action bit through the reinforcement learning scheduling strategy, write the complete set shipment increment into the benefit word, connect the single-end cache occupancy bit newly added after the resource scheduling column is executed to the repair occupancy bit to form the loss word, and use the benefit word to subtract the loss word to obtain the action reward of this round. S5-3. Use the original action value corresponding to the scheduling action bit to access the action reward of this round and the action value of the next event bit to generate an updated action value. Select the corresponding order segment write depth according to the bit order result of the updated action value and output the target order segment write depth.

Images