Method and system for dynamic adjustment of safety stock of spare parts based on demand fluctuation prediction

Abstract

The application discloses a spare part safety inventory dynamic adjustment method and system based on demand fluctuation prediction, and relates to the technical field of inventory prediction. The spare part safety inventory dynamic adjustment method based on demand fluctuation prediction comprises the following steps: S1, collecting hardware maintenance demand data, inventory flow data, fault point word table data and historical safety inventory records, and performing standardization and normalization processing; S2, performing demand window grid credible analysis according to the matching relationship and path length of the confirmation event code and the delivery event code; S3, constructing a fault spare part demand fluctuation prediction model based on the window grid credible value, the demand occurrence window table and the inventory flow data; and S4, constructing an inventory trigger flow chart based on the fluctuation passing mark, the spare part demand fluctuation value, the safety inventory adjustment record and the inventory flow data. The problems of replenishment trigger lag, advance mis-triggering and safety inventory repeated change caused by the inconsistency between the demand fluctuation window and the real spare part demand occurrence window are solved.

Description

Technical Field

[0001] This invention relates to the field of inventory forecasting technology, specifically to a method and system for dynamically adjusting spare parts safety stock based on demand fluctuation forecasting. Background Technology

[0002] In computer hardware repair, spare parts inventory management requires statistical analysis based on equipment brand, model, batch, fault type, and spare parts outbound records. Repair shops or service platforms typically record repair order numbers, pick-up times, initial inspection times, fault confirmation times, repair completion times, spare parts numbers, spare parts names, outbound quantities, available inventory quantities, replenishment times, and arrival times. They then determine safety stock thresholds for spare parts based on historical repair demand and inventory flow. For commonly used spare parts such as screens, batteries, keyboards, motherboards, hard drives, fans, interface boards, and ribbon cables, the appropriateness of safety stock settings directly impacts repair response speed, spare parts turnover efficiency, and inventory occupancy levels. Existing spare parts safety stock management methods typically forecast demand based on historical outbound quantities, remaining inventory, replenishment cycles, and arrival cycles, and generate replenishment recommendations by combining average consumption, consumption fluctuations, out-of-stock records, and safety stock thresholds. Some solutions also group historical repair records by equipment brand, model, fault name, and spare parts number to achieve differentiated inventory management for different hardware categories. In computer repair scenarios, different brands, models, and batches of equipment may have different areas of concentrated failure. Simply relying on spare parts names or outbound quantities to count needs makes it difficult to accurately reflect the impact of changes in failure points on the pace of spare parts consumption.

[0003] For example, the invention patent with announcement number CN116188061B discloses a method, device, electronic device, and storage medium for predicting commodity sales. This method successively introduces a sales increment prediction model corresponding to each factor and a comprehensive sales prediction model, comprehensively considering the impact of each factor on commodity sales. This can improve prediction accuracy, make prediction results more accurate, and enhance volatility tracking, thus effectively solving the problems of low prediction accuracy and weak volatility tracking under the influence of different factors. It has strong versatility and can be applied to the sales prediction of various categories of commodities in the general retail industry, such as electronic products, food and groceries, and fresh produce. It overcomes the shortcomings of tree models, which cannot provide prediction values ​​exceeding the upper and lower limits of historical actual sales, while also having the advantages of utilizing the characteristics of each factor and considering the nonlinear relationships between prediction variables.

[0004] For example, the invention patent with announcement number CN120822660B discloses an intelligent warehouse dynamic optimization system based on digital twin technology, including the following steps: S1: Creating a digital twin model: acquiring basic data of the physical warehouse and constructing an accurate digital model; S2: Data collection and integration: deploying IoT devices and realizing real-time data transmission and processing; S3: Constructing analysis algorithms: developing advanced analysis algorithms and implementing machine learning models; S4: Dynamic optimization and adjustment: automatically adjusting the storage layout and optimizing the inbound and outbound processes; S5: Prediction and early warning: predicting potential problems and setting alarm mechanisms based on historical data analysis, significantly improving data real-time performance and accuracy, enhancing environmental monitoring capabilities and ensuring cargo safety, and greatly improving space utilization.

[0005] However, in actual maintenance processes, a single maintenance record typically involves multiple stages, including receiving the machine, initial inspection, fault confirmation, spare parts issuance, and maintenance completion. Existing demand forecasting methods often directly use the machine reception time or spare parts issuance time as the demand occurrence time to construct a time series and estimate spare parts demand fluctuations. When using the machine reception time, the reception record may only contain customer descriptions or preliminary fault symptoms; the fault point and corresponding spare parts have not yet been confirmed, easily leading to the inclusion of unconfirmed records in demand statistics. When using the spare parts issuance time, the issuance action usually occurs after inspection and confirmation, delaying the actual demand signal until the inventory deduction point, causing a shift in the peak demand fluctuation on the timeline. Regarding replenishment triggers, existing safety stock adjustment methods mostly use situations where the available inventory falls below the safety stock threshold, the number of spare parts issued increases, or the historical consumption average rises as trigger criteria. Since there are different period spans between fault confirmation time, spare parts issuance time, replenishment order generation time, and replenishment arrival time, starting the replenishment trigger too early can easily lead to inventory backlog, while starting it too late can easily lead to stockout risks. Frequent adjustments to the safety stock threshold within adjacent statistical periods can also lead to unstable replenishment rhythms, affecting the feasibility of inventory management results.

[0006] Therefore, in order to address the above issues, there is an urgent need for a method and system for dynamically adjusting spare parts safety stock based on demand fluctuation forecasts. Summary of the Invention

[0007] Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a method and system for dynamically adjusting spare parts safety stock based on demand fluctuation forecasting. This solves the problems of delayed replenishment triggering, premature false triggering, and repeated changes in safety stock caused by the inconsistency between the demand fluctuation window and the actual spare parts demand window.

[0009] Technical solution

[0010] To achieve the above objectives, this invention provides the following technical solution: a method for dynamically adjusting spare parts safety stock based on demand fluctuation forecasting, comprising: S1, collecting hardware repair demand data, inventory flow data, fault point vocabulary data, and historical safety stock records, and performing standardization and normalization processing; S2, performing demand window grid reliability analysis based on the matching relationship and path length of confirmation event codes and outbound event codes, and performing demand occurrence cycle confirmation and replenishment candidate blocking based on the grid reliability analysis results; S3, constructing a fault spare parts demand fluctuation forecasting model based on window grid reliability values, demand occurrence window tables, and inventory flow data, and performing fluctuation through-marking and abnormal fluctuation sample recording; S4, constructing an inventory trigger flow graph based on fluctuation through-marking, spare parts demand fluctuation values, safety stock adjustment records, and inventory flow data, and performing trigger starting point migration and path rearrangement.

[0011] Furthermore, the specific steps for collecting hardware repair demand data, inventory flow data, fault point terminology data, and historical safety stock records are as follows: Collecting hardware repair demand data: Retrieving the repair order number, equipment brand, equipment model, equipment batch, original fault name, arrival time, initial inspection time, fault confirmation time, and repair completion time from the repair work order; Collecting inventory flow data: Retrieving the spare part number, spare part name, outbound time, outbound quantity, available inventory quantity, current period safety stock threshold, replenishment order generation time, and replenishment arrival time from the inventory flow; Collecting fault point terminology data: Retrieving the standard fault point name, synonyms, prohibited terms, and fault point-related spare part numbers from the fault point terminology; Collecting historical safety stock records: Retrieving the spare part number, previous period safety stock threshold, and replenishment trigger point from the historical safety stock records.

[0012] Further, the specific steps for standardization and normalization are as follows: Hardware repair demand data, inventory flow data, fault point thesaurus data, and historical safety stock records are cleaned, deduplicated, time-standardized, and field-normalized to generate a unified cycle number; a replenishment generation mark is generated based on the replenishment order generation time, and an arrival backfill mark is generated based on the replenishment arrival time; the original fault point names are mapped to standard fault point names to generate standardized fault point records; fault grouping identifiers are generated using equipment brand, equipment model, equipment batch, and standard fault point names; records where the current cycle safety stock threshold is greater than the previous cycle safety stock threshold are written to the threshold increase mark, and records where the current cycle safety stock threshold is less than the previous cycle safety stock threshold are written to the threshold decrease mark. A time chain record is generated based on the maintenance order number, cycle number, receiving time, initial inspection time, fault confirmation time, outbound time, and maintenance completion time, and the receiving event code, initial inspection event code, confirmation event code, outbound event code, and completion event code are written into it; a confirmation waiting segment duration is generated based on the time interval between the fault confirmation time and the outbound time; candidate fields for demand occurrence cycle and consumption verification cycle are generated based on the confirmation event code and the outbound event code; a replenishment order generation cycle is generated based on the replenishment order generation time, and a replenishment arrival cycle is generated based on the replenishment arrival time; the difference in available inventory quantity before and after the window is generated based on the available inventory quantity of adjacent cycles; and a faulty spare parts index table is generated based on the fault grouping identifier, standard fault point name, fault point associated spare parts number, and spare parts number.

[0013] Furthermore, the specific steps for demand window grid reliability analysis based on the matching relationship and path length of the confirmation event code and the outbound event code are as follows: A window path matrix is ​​constructed using the fault group identifier and maintenance order number as a combined row index and the cycle number as a column index; the receiving event code, initial inspection event code, confirmation event code, outbound event code, and completion event code are written into the window path matrix; the path length between the confirmation event code and the outbound event code along the same maintenance order number is read, and the matching relationship between the standard fault point name corresponding to the confirmation event code and the spare part number in the fault spare part index table is verified; when the confirmation event code is located before the outbound event code, and the spare part number matching relationship is valid, the confirmation event code is... The grid segment between the code and the outbound event code is written into a candidate closed path; when the path length of the candidate closed path does not exceed the window span threshold, the candidate closed path is written into the closed path marker, and the corresponding window is written into the trusted window queue; when the path length of the candidate closed path exceeds the window span threshold, the candidate closed path is written into the drift span marker, and the corresponding window is written into the drift window queue; when there is no acknowledgment event code between the receiving event code and the outbound event code, the corresponding grid segment is written into the noise hole marker, and the corresponding window is written into the noise window queue; based on the coverage position of the closed path marker, the noise hole marker, and the drift span marker under the same fault group identifier, the window grid trusted value is obtained.

[0014] Furthermore, the specific steps for confirming the demand occurrence cycle and blocking replenishment candidates based on the grid reliability analysis results are as follows: By comparing the window grid reliability value and reliability threshold in real time, when the window grid reliability value is greater than or equal to the reliability threshold, the corresponding grid segment is written into the drift window queue, the cycle number where the confirmation event code is located is determined as the demand occurrence cycle, and the corresponding fault group identifier and spare part number are input into the faulty spare part demand fluctuation prediction model; when the window grid reliability value is less than the reliability threshold, diversion processing is performed according to the path marker type; if the drift span marker covers the grid segment between the confirmation event code and the outbound event code, a demand occurrence window table is created with the cycle number where the confirmation event code is located, and the cycle number where the outbound event code is located is written as the consumption verification cycle; if the closed path marker is missing, the safety stock threshold of the previous cycle is retained, and the corresponding grid segment is written into the window rearrangement list.

[0015] Furthermore, the specific steps for constructing a faulty spare parts demand fluctuation prediction model based on window grid reliability values, demand occurrence window tables, and inventory flow data are as follows: Using window grid reliability values, demand occurrence window tables, consumption verification cycles, standardized fault point records, and inventory flow data as basic inputs, the cycle numbers of the confirmation event codes and the outbound event codes are associated and paired according to the maintenance order number. A time-series bucket coding method is used to convert the cycle number difference between the confirmation event code and the outbound event code into a time-delay bucket sequence. The confirmation waiting segment duration and the difference in available inventory quantity before and after the window are written into the feature matrix. When a confirmation event code in the demand occurrence window table forms an outbound event code matching the spare parts number within the corresponding consumption verification cycle, the fluctuation label value is written as a valid fluctuation marker. When a confirmation event code does not form an outbound event code matching the spare parts number within the corresponding consumption verification cycle, the fluctuation label value is written as a non-fluctuation marker. A gradient boosting decision tree method is used to construct a faulty spare parts demand fluctuation prediction model based on the feature matrix and fluctuation label values, and the spare parts demand fluctuation value is output.

[0016] Furthermore, the specific steps for executing the fluctuation through-marking and abnormal fluctuation sample recording are as follows: By comparing the spare parts demand fluctuation value and the fluctuation threshold in real time, when the spare parts demand fluctuation value is greater than or equal to the fluctuation threshold, a fluctuation through mark is generated and written into the safety stock adjustment record, with the cycle number of the outbound event code as the consumption verification cycle; when the spare parts demand fluctuation value is less than the fluctuation threshold, a fluctuation pending verification mark is generated, and the safety stock threshold of the previous cycle is retained; when an outbound event code matching the spare parts number appears subsequently, the newly added confirmation event code, outbound event code, time lag bucket sequence, and fluctuation label value are written back to the faulty spare parts demand fluctuation prediction model. If the re-output spare parts demand fluctuation value is still less than the fluctuation threshold, the corresponding sample is written into the abnormal fluctuation sample record.

[0017] Furthermore, the specific steps for constructing an inventory trigger flow diagram based on fluctuation pass markers, spare parts demand fluctuation values, safety stock adjustment records, and inventory flow data are as follows: Construct an inventory trigger flow diagram based on fluctuation pass markers, spare parts demand fluctuation values, safety stock adjustment records, available inventory quantity, current period safety stock threshold, demand occurrence cycle, consumption verification cycle, replenishment order generation cycle, and replenishment arrival cycle; Based on the demand occurrence cycle, records where spare parts demand fluctuation values ​​are greater than or equal to the fluctuation threshold, consumption verification cycle, replenishment order generation cycle, and replenishment arrival cycle, generate demand window write nodes, fluctuation pass nodes, inventory deduction nodes, replenishment generation nodes, and arrival backfill nodes, respectively; use spare parts number as the row index and cycle number. An inventory level trajectory table is constructed for the column index, and the three trigger paths, with the demand window write node, fluctuation pass node, and inventory deduction node as the replenishment starting points, are replayed to the inventory level trajectory table. When a grid appears before the replenishment arrival cycle where the available inventory quantity is less than the safety stock threshold of the current cycle, a delayed trigger flag is written. When a grid appears after the replenishment arrival cycle where the available inventory quantity minus the outbound quantity is still greater than the safety stock threshold of the current cycle, an early trigger flag is written. When threshold adjustment flags and threshold reduction flags are written alternately in adjacent cycles, a threshold swing flag is written. Based on the coverage quantity and coverage position of the delayed trigger flag, early trigger flag, and threshold swing flag in the arrangement of the three trigger paths, the inventory trajectory fit value is obtained.

[0018] Further, the specific steps for executing the trigger point migration and path rearrangement are as follows: By comparing the inventory trajectory alignment value and alignment threshold in real time, when the inventory trajectory alignment value is greater than or equal to the alignment threshold, the trigger path with the fewest covered quantities of lagging trigger markers, early trigger markers, and threshold swing markers is selected, and the selected trigger path is written into the safety stock adjustment record to generate a replenishment order instruction; when the inventory trajectory alignment value is less than the alignment threshold, the lagging trigger markers, early trigger markers, and threshold swing markers in the arrangement of the three trigger paths are read; for trigger paths carrying lagging trigger markers, the replenishment trigger point is moved forward from the inventory deduction node to the fluctuation passage node, if the fluctuation... If a missing part grid still exists in the corresponding grid of the active node, the process is moved forward to the demand window write node, and a replenishment trigger record is generated according to the demand occurrence cycle. For trigger paths with an early trigger flag, the replenishment trigger starting point is moved from the demand window write node to the fluctuation pass node. If there are still overstock grids after the replenishment arrival cycle, the process is moved to the inventory deduction node, and a replenishment trigger record is generated according to the consumption verification cycle. For trigger paths with a threshold swing flag, the current cycle safety stock threshold writing is stopped, the previous cycle safety stock threshold is used, and the corresponding fault group identifier, spare part number, replenishment trigger starting point, and flag type are written into the safety stock adjustment record.

[0019] Furthermore, the second aspect of this invention provides a dynamic adjustment system for spare parts safety stock based on demand fluctuation forecasting, and applies a method for dynamic adjustment of spare parts safety stock based on demand fluctuation forecasting, including: a maintenance timeline acquisition and preprocessing module, used to collect hardware maintenance demand data, inventory flow data, fault point vocabulary data, and historical safety stock records, and perform standardization and normalization processing; a demand window grid reliability analysis module, used to perform demand window grid reliability analysis based on the matching relationship and path length of confirmation event codes and outbound event codes, and perform demand occurrence cycle confirmation and replenishment candidate blocking based on the grid reliability analysis results; a faulty spare parts demand fluctuation forecasting model construction module, used to construct a faulty spare parts demand fluctuation forecasting model based on window grid reliability values, demand occurrence window tables, and inventory flow data, and perform fluctuation through-mark writing and abnormal fluctuation sample recording; and a safety stock trigger trajectory correction module, used to construct an inventory trigger flow diagram based on fluctuation through-marks, spare parts demand fluctuation values, safety stock adjustment records, and inventory flow data, and perform trigger starting point migration and path rearrangement.

[0020] The present invention has the following beneficial effects: (1) This invention converts the periodic number difference between the confirmation event code and the outbound event code into a time-delay bucket sequence by using a time-series bucket coding method. It also constructs a fault spare parts demand fluctuation prediction model by combining the confirmation waiting segment duration, the difference in available inventory quantity before and after the window, and the fluctuation label value. This enables the connection between the fault point fluctuation and the subsequent spare parts outbound to be learned and output by the feature matrix, thereby achieving the machine learning prediction effect for the time delay relationship of fault spare parts. This effectively solves the problem in the prior art that it is difficult to identify the variable time delay relationship between fault confirmation and spare parts consumption by judging spare parts demand based solely on the outbound quantity fluctuation.

[0021] (2) In this invention, the receiving event code, initial inspection event code, confirmation event code, outbound event code and completion event code are recorded through the window path matrix. The window grid reliability value is obtained based on the matching relationship between the confirmation event code and the outbound event code and the path length. This allows the demand occurrence cycle to be separated from the receiving noise and outbound lag, thereby achieving the effect of real demand window calibration. This effectively solves the problem of fluctuation peak shift caused by directly using the receiving time or outbound time to construct the demand sequence in the prior art.

[0022] (3) In this invention, by comparing the fluctuation value of spare parts demand with the fluctuation threshold, the fault group identifier and spare parts number that reach the fluctuation threshold are written into the replenishment candidate queue, and the samples that do not reach the fluctuation threshold are written into the abnormal fluctuation sample record, so that the generation of replenishment candidates is constrained by the fault spare parts acceptance relationship, thereby achieving the effect of reliable screening of replenishment candidates, and effectively solving the problem in the prior art that the fluctuation of the fault point does not lead to the consumption of spare parts but still triggers the replenishment candidate.

[0023] (4) This invention uses the inventory trigger flow diagram and inventory level trajectory table playback to create three trigger paths with the demand window write node, the receiving through node and the inventory deduction node as the replenishment starting point. Based on the delayed trigger mark, the early trigger mark and the threshold swing mark, the inventory trajectory fit value is obtained, so that the replenishment trigger starting point can be migrated and rearranged according to the changes in inventory level. This achieves the effect of correcting the safety stock replenishment rhythm and effectively solves the problems of delayed replenishment trigger, early false trigger and repeated changes in safety stock threshold in the prior art.

[0024] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0025] Figure 1 This is a flowchart of the method for dynamically adjusting spare parts safety stock based on demand fluctuation forecasting according to the present invention. Figure 2 This is a structural diagram of the spare parts safety stock dynamic adjustment system based on demand fluctuation forecasting of the present invention; Figure 3 This is a schematic diagram of the maintenance event code window path matrix of the present invention; Figure 4 This is a comparison chart of inventory level trajectories under different replenishment trigger paths according to the present invention. Detailed Implementation

[0026] 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.

[0027] Please see Figures 1-4 This invention provides a technical solution: a method for dynamically adjusting spare parts safety stock based on demand fluctuation forecasting, comprising: S1, collecting hardware repair demand data, inventory flow data, fault point vocabulary data, and historical safety stock records, and performing standardization and normalization processing; S2, performing demand window grid reliability analysis based on the matching relationship and path length of confirmation event codes and outbound event codes, and performing demand occurrence cycle confirmation and replenishment candidate blocking based on the grid reliability analysis results; S3, constructing a fault spare parts demand fluctuation forecasting model based on window grid reliability values, demand occurrence window tables, and inventory flow data, and performing fluctuation through-marking writing and abnormal fluctuation sample recording; S4, constructing an inventory trigger flow graph based on fluctuation through-marking, spare parts demand fluctuation values, safety stock adjustment records, and inventory flow data, and performing trigger starting point migration and path rearrangement.

[0028] Specifically, the steps for collecting hardware repair demand data, inventory flow data, fault point thesaurus data, and historical safety stock records are as follows: Collect hardware repair demand data: Read the repair order number, equipment brand, equipment model, equipment batch, original fault name, arrival time, initial inspection time, fault confirmation time, and repair completion time from the repair work order. Use the repair order number as the primary key for associating each time field and fault point record within the same hardware repair process. This allows for subsequent tracking of the sequence of arrival event codes, initial inspection event codes, confirmation event codes, outbound event codes, and completion event codes based on the same repair order number. Relationships; Inventory flow data collection: Read spare part number, spare part name, outbound time, outbound quantity, available inventory quantity, current period safety stock threshold, replenishment order generation time, and replenishment arrival time from the inventory flow data. The current period safety stock threshold is derived from the safety stock threshold field for the same spare part number within the current period. If a safety stock adjustment record has been generated for the current period, the safety stock threshold written in the safety stock adjustment record is used as the current period safety stock threshold. Fault point terminology data collection: Read standard fault point names, synonyms, disabled terms, and fault points from the fault point terminology. The associated spare parts number is established and maintained based on the standard fault point name, equipment brand, equipment model, equipment batch, replacement spare parts number in historical maintenance completion records, and outbound spare parts number in inventory flow. When the same standard fault point name corresponds to multiple spare parts numbers, multiple candidate associated records are retained in the fault spare parts index table. A matching priority is generated according to the following order: consistent equipment brand, consistent equipment model, consistent equipment batch, historical matching count, and recent outbound count, ensuring that the standard fault point name corresponding to the confirmation event code can be matched with the spare parts number in the fault spare parts index table. (Data collection) Historical safety stock records: Read the spare part number, the safety stock threshold of the previous period, and the replenishment trigger start point from the historical safety stock records. The safety stock threshold of the previous period comes from the safety stock threshold field written at the end of the previous period for the same spare part number. The replenishment trigger start point is used to record the initial node of the replenishment trigger path. Allowed values ​​include the demand window write node, the fluctuation pass node, and the inventory deduction node. The initial replenishment trigger start point is determined according to the selected trigger path in the safety stock adjustment record of the previous period. If there is no selected trigger path in the previous period, the inventory deduction node is used as the initial replenishment trigger start point.

[0029] In this implementation plan, a unified association link is established through maintenance order number, spare part number, cycle number, and fault group identifier, enabling hardware maintenance demand data, inventory flow data, fault point thesaurus data, and historical safety stock records to form a traceable data foundation. The current cycle safety stock threshold and the previous cycle safety stock threshold are provided by inventory flow data and historical safety stock records, respectively. The fault point associated spare part number forms a candidate association based on the standard fault point name, equipment brand, equipment model, equipment batch, and historical replacement records. The replenishment trigger point is limited to the demand window write node, fluctuation pass node, and inventory deduction node, thereby providing data support with consistent field sources, clear matching relationships, and traceable trigger paths for subsequent window grid reliability analysis, fault spare part demand fluctuation prediction model construction, and inventory trigger flow diagram generation.

[0030] Specifically, the standardization and normalization processes are as follows: Hardware repair request data, inventory flow data, fault point terminology data, and historical safety stock records are cleaned, deduplicated, time-standardized, and field-normalized. During cleaning, records lacking repair order numbers, spare parts numbers, equipment brands, equipment models, or the original name of the fault point are deleted. Records with outbound times earlier than receiving times, fault confirmation times earlier than initial inspection times, repair completion times earlier than fault confirmation times, and replenishment arrival times earlier than replenishment order generation times are also deleted. During deduplication, the repair order number, equipment brand, equipment model, equipment batch, and the original name of the fault point are used as the duplicate repair order detection keys, retaining records with later fault confirmation times for future reference. Part number, outbound time, and outbound quantity are used as duplicate keys for inventory flow determination, retaining records written later. For time standardization, the starting point of the business day in the inventory management terminal is used as the cycle number, with the calendar day as the cycle granularity. The receiving time, initial inspection time, fault confirmation time, outbound time, repair completion time, replenishment order generation time, and replenishment arrival time are mapped to the same calendar day cycle. The cycle number increments by one digit for each calendar day, generating a unified cycle number. A replenishment generation flag is generated based on the cycle number of the replenishment order generation time, and an arrival backfill flag is generated based on the cycle number of the replenishment arrival time. Field normalization is performed on equipment brand, equipment model, equipment batch, and spare part number, removing spaces. After checking for differences in connectors and case sensitivity, the data is written into a unified encoding field. The original fault point name is segmented, disabled for deletion, synonyms are merged, and mapped to a standard fault point name to generate standardized fault point records. Records that do not match a standard fault point name are written to a fault point mapping list and are not yet included in fault grouping identifier generation. Fault grouping identifiers are generated using equipment brand, equipment model, equipment batch, and standard fault point name. The safety stock threshold for the previous period is read from historical safety stock records, and the safety stock threshold for the current period is read from inventory flow data. The current period safety stock threshold is written after the end of the period in which the replenishment order is generated. Records where the current period safety stock threshold is greater than the previous period safety stock threshold are written to the threshold. The value is adjusted upwards, and records where the current period's safety stock threshold is less than the previous period's safety stock threshold are marked downwards. A time chain record is generated based on the repair order number, cycle number, receiving time, initial inspection time, fault confirmation time, outbound time, and repair completion time, and the receiving event code, initial inspection event code, confirmation event code, outbound event code, and completion event code are written. The confirmation waiting segment duration is generated based on the time interval between the fault confirmation time and the outbound time. The confirmation waiting segment duration is expressed in terms of natural day cycles. When the fault confirmation time and the outbound time are in the same natural day cycle, it is recorded as one cycle. When they are in different natural day cycles, it is calculated by adding one to the difference in cycle number. Records corresponding to abnormal timestamps are written to the time abnormality list.Based on the cycle number of the confirmation event code, a candidate field for the demand occurrence cycle is generated. This field represents the candidate cycle for the spare parts demand window after the fault point has been detected and confirmed. Based on the cycle number of the outbound event code, a candidate field for the consumption verification cycle is generated. This field represents the candidate cycle for the actual outbound delivery of spare parts and the verification of demand fulfillment. Both need to be written into the demand occurrence window table and the consumption verification cycle table after the subsequent closed path marker is established. A replenishment order generation cycle is generated based on the replenishment order generation time, and a replenishment arrival cycle is generated based on the replenishment arrival time. The cycle number of the confirmation event code is used as the starting point of the window, and the outbound... The period number of the event code is used as the end point of the window. The available inventory quantity in the period before the start of the window and the available inventory quantity in the period after the end of the window are read, and the difference between the two is used as the difference in available inventory quantity before and after the window. A faulty spare parts index table is generated based on the fault group identifier, standard fault point name, fault point associated spare parts number, and spare parts number. The faulty spare parts index table uses the fault group identifier, standard fault point name, and spare parts number as unique keys. Duplicate key records retain the record corresponding to the most recent outbound time. One-to-many mappings are arranged from most to least frequent spare parts outbounds, while many-to-one mappings are split into multiple index items based on equipment brand, equipment model, and equipment batch.

[0031] This implementation plan cleanses, deduplicates, standardizes time data, and normalizes fields in hardware maintenance demand data, inventory flow data, fault point thesaurus data, and historical safety stock records. This ensures that maintenance order number, spare part number, equipment brand, equipment model, equipment batch, standard fault point name, arrival time, initial inspection time, fault confirmation time, outbound time, maintenance completion time, replenishment order generation time, and replenishment arrival time are all entered into a unified periodic numbering caliber. By generating standardized fault point records, fault grouping identifiers, time chain records, arrival event codes, initial inspection event codes, confirmation event codes, outbound event codes, completion event codes, confirmation waiting segment duration, candidate fields for demand occurrence cycle, candidate fields for consumption verification cycle, replenishment order generation cycle, replenishment arrival cycle, difference in available inventory quantity before and after the window, and a faulty spare part index table, subsequent window path matrix, window grid confidence value, time lag bucket sequence, and spare part demand fluctuation value have a consistent data source. This reduces the interference of inconsistent fault point names, inconsistent time field calibers, spare part number mapping conflicts, and inconsistent inventory threshold cycles on demand fluctuation prediction results.

[0032] Specifically, the steps for demand window grid reliability analysis based on the matching relationship between confirmation event codes and outbound event codes and path length are as follows: A window path matrix is ​​constructed using the fault group identifier and maintenance order number as a combined row index and the cycle number as a column index. Each grid cell in the window path matrix records data in an event code list format. The event code list is arranged according to the business sequence of arrival event code, initial inspection event code, confirmation event code, outbound event code, and completion event code. When multiple event codes are written within the same cycle number, they are appended according to the above business sequence. When the same event code is written repeatedly, the later event code record is retained. In a given week... When no event code is generated within a period number, a null value marker is written to the corresponding grid cell; the receiving event code, initial inspection event code, confirmation event code, outbound event code, and completion event code are written to the window path matrix; the path length between the confirmation event code and the outbound event code is read along the same maintenance order number. The path length is the difference between the period number of the outbound event code and the period number of the confirmation event code plus one, and the matching relationship between the standard fault point name corresponding to the confirmation event code and the spare part number in the fault spare part index table is verified; a matching relationship is established when there is an associated record in the fault spare part index table where the standard fault point name and the spare part number are consistent. If the same standard fault point name... For multiple spare part numbers, the matching priority is determined according to the following order: consistent equipment brand, consistent equipment model, consistent equipment batch, spare part number associated with the fault point, and number of hits in historical maintenance completion records. The spare part number with the highest priority is used as the matching spare part number. When the confirmation event code precedes the outbound event code and the spare part number matching relationship is established, the grid segment between the confirmation event code and the outbound event code is written into the candidate closed path. The window span threshold is derived from the statistical results of the periodic number span between the confirmation event code and the outbound event code in the historical closed path sample. It is determined by adding one quartile interval to the median of the periodic number span in the historical closed path sample. The range is from 1 to 12 period numbers; when the path length of a candidate closed path does not exceed the window span threshold, the candidate closed path is written into the closed path marker, and the corresponding window is written into the trusted window queue; when the path length of a candidate closed path exceeds the window span threshold, the candidate closed path is written into the drift span marker, and the corresponding window is written into the drift window queue; when there is a missing confirmation event code between the receiving event code and the outbound event code, the corresponding grid segment is written into the noise hole marker, and the corresponding window is written into the noise window queue; based on the coverage position of the closed path marker, the noise hole marker, and the drift span marker under the same fault group identifier, the window grid trusted value is obtained.

[0033] Table 1 shows the data table for the routing determination of the window path matrix of maintenance event codes in this embodiment. The table uses the fault group identifier and maintenance order number as the joint row index to record the cycle number of the receiving event code, initial inspection event code, confirmation event code, outbound event code and completion event code. The routing result in the window path matrix is ​​represented by the matching mark, path mark code and queue code. The matching mark of 1 indicates that the spare part number matching relationship is established, and the matching mark of 0 indicates that the spare part number matching relationship is not formed. The cycle number of the confirmation event code is 0, which is a null mark. The path length is 0, which means that there is no confirmation event code between the receiving event code and the outbound event code and it does not participate in the candidate closed path determination. The path mark codes 1, 2 and 3 correspond to the closed path mark, drift span mark and noise hole mark, respectively. The queue codes 1, 2 and 3 correspond to the reliable window queue, drift window queue and noise window queue, respectively.

[0034] The fault group identifier of record 1 is FG01, the repair order number is 001, the cycle number of the receiving event code is 1, the cycle number of the initial inspection event code is 2, the cycle number of the confirmation event code is 3, the cycle number of the outbound event code is 4, the cycle number of the completion event code is 5, the path length is 2, the matching mark is 1, and the determined path mark code is 1, and the queue code is 1; The fault group identifier of record 2 is FG01, the repair order number is 002, the cycle number of the confirmation event code is 4, the cycle number of the outbound event code is 5, the path length is 2, the matching mark is 1, and the determined path mark code is 1, and the queue code is 1; The fault group identifier of record 3 is FG02, the repair order number is 003, the cycle number of the confirmation event code is 3, the cycle number of the outbound event code is 7, the path length is 5, and the matching mark is... 1. The path marker code and queue code are determined to be 2. 2. The fault group identifier of record 4 is FG02, the maintenance order number is 004, the cycle number of the confirmation event code is 3, the cycle number of the outbound event code is 6, the path length is 4, and the matching marker is 1. The path marker code and queue code are determined to be 2. 3. The fault group identifier of record 5 is FG03, the maintenance order number is 005, the cycle number of the confirmation event code is 0, the cycle number of the outbound event code is 5, the path length is 0, and the matching marker is 0. The path marker code and queue code are determined to be 3. 4. The fault group identifier of record 6 is FG03, the maintenance order number is 006, the cycle number of the confirmation event code is 0, the cycle number of the outbound event code is 6, the path length is 0, and the matching marker is 0. The path marker code and queue code are determined to be 3.

[0035] Table 1. Maintenance Event Code Window Path Matrix Triage Judgment Data Table

[0036]

[0037] like Figure 3 The diagram shown is a schematic of the maintenance event code window path matrix in this embodiment. The horizontal axis represents the cycle number, and the vertical axis represents the fault group identifier and maintenance order number. The window path matrix uses the fault group identifier and maintenance order number as a combined row index and the cycle number as a column index. Figure 3 As shown in Table 1, the motherboard power supply repair order 001 corresponds to FG01 and 001 in Table 1, the motherboard power supply repair order 002 corresponds to FG01 and 002 in Table 1, the screen display repair order 003 corresponds to FG02 and 003 in Table 1, the screen display repair order 004 corresponds to FG02 and 004 in Table 1, the hard drive recognition repair order 005 corresponds to FG03 and 005 in Table 1, and the hard drive recognition repair order 006 corresponds to FG03 and 006 in Table 1; different colors in the figure represent connections... The event codes are: machine event code, initial inspection event code, confirmation event code, outbound event code, and completion event code. Closed path markers indicate a candidate closed path between the confirmation event code and the outbound event code. Drift span markers indicate that the window span threshold is crossed between the confirmation event code and the outbound event code. Noise hole markers indicate the absence of a confirmation event code between the receiving event code and the outbound event code. This demonstrates the basis for the formation of the reliable window queue, drift window queue, and noise window queue, providing a rasterized example for obtaining the window grid's reliability value. Combined with... Figure 3 As shown in Table 1, different colors are used to identify the receiving event code, initial inspection event code, confirmation event code, outbound event code, and completion event code. The distribution of each event code on the cycle number is displayed in the corresponding rows of the motherboard power supply repair order 001, motherboard power supply repair order 002, screen display repair order 003, screen display repair order 004, hard disk recognition repair order 005, and hard disk recognition repair order 006. The closed path marker in the figure indicates that a candidate closed path is formed between the confirmation event code and the outbound event code. The drift span marker indicates that multiple cycle numbers are crossed between the confirmation event code and the outbound event code. The noise hole marker indicates that there is no confirmation event code between the receiving event code and the outbound event code. This shows the basis for the formation of the reliable window queue, drift window queue, and noise window queue, and provides a rasterized example for obtaining the window grid reliability value.

[0038] In this implementation scheme, the receiving event code, initial inspection event code, confirmation event code, outbound event code, and completion event code under the same fault group identifier and maintenance order number are uniformly mapped to the cycle number grid through a window path matrix. The integrity of the matrix unit data structure is ensured by using an event code list, null value marking, and duplicate event retention rules. By using the path length caliber, the matching relationship of the faulty spare parts index table, and the window span threshold, the grid segment between the confirmation event code and the outbound event code is closed, drifted, and noise-distinguished, so that the reliable window queue, drift window queue, and noise window queue have clear data sources. This enables the identification of unconfirmed noise caused by the receiving event code and cycle shift caused by the outbound event code, providing a traceable and reliable window grid basis for subsequent demand cycle confirmation, input of faulty spare parts demand fluctuation prediction model input, and replenishment candidate blocking.

[0039] Specifically, the steps for confirming demand occurrence cycles and blocking replenishment candidates based on the grid reliability analysis results are as follows: By comparing the window grid reliability value and reliability threshold in real time, when the window grid reliability value is greater than or equal to the reliability threshold, it indicates that a usable closed relationship has been formed between the confirmation event code, the outbound event code, and the spare part number. The corresponding grid segment is written into the demand occurrence window table, the cycle number where the confirmation event code is located is determined as the demand occurrence cycle, and the corresponding fault group identifier and spare part number are input into the faulty spare part demand fluctuation prediction model. When the window grid reliability value is less than the reliability threshold, diversion processing is performed according to the path marking type. If the drift span mark covers the grid segment between the confirmation event code and the outbound event code, a demand occurrence window table is created with the cycle number where the confirmation event code is located, and the cycle number where the outbound event code is located is written as the consumption verification cycle. During this period, the outbound event code is used only as the basis for fulfilling spare parts consumption. If the closed path marker is missing, the closed path marker is searched within the corresponding grid segment from the same fault group identifier, the same maintenance order number, and the current confirmation event code to the outbound event code. When the search results do not show a matching relationship between the confirmation event code and the spare parts number, the safety stock threshold of the previous period is retained, and the corresponding grid segment is written into the window rearrangement list. The window rearrangement list records at least the fault group identifier, spare parts number, maintenance order number, original window start point, original window end point, rearrangement reason, and pending processing marker. The rearrangement reason includes missing closed path marker, spare parts number matching failure, and path length exceeding the window span threshold. The pending processing marker is used to indicate that the corresponding grid segment needs to participate in the window grid reliability value acquisition again after the confirmation event code, outbound event code, or spare parts number matching relationship is subsequently added.

[0040] In this implementation scheme, by comparing the window grid confidence value with the confidence threshold, grid segments with a closed relationship between the confirmation event code, the outbound event code, and the spare part number are written into the demand occurrence window table. The demand occurrence cycle is determined by the cycle number of the confirmation event code, so that the fault group identifier and the spare part number can be used as effective inputs for the fault spare part demand fluctuation prediction model. For grid segments with insufficient window grid confidence values, they are diverted according to the missing drift span marker and closed path marker. The outbound event code is limited to the consumption verification cycle to avoid the outbound time replacing the demand occurrence cycle. The fault group identifier, spare part number, maintenance order number, original window start point, original window end point, rearrangement reason, and pending processing marker are recorded through the window rearrangement list. This ensures that grid segments with missing closed path markers, failed spare part number matching, and path lengths exceeding the window span threshold do not trigger safety stock threshold changes, thereby reducing the interference of receiving event code noise and outbound event code lag on the judgment of spare part demand fluctuations.

[0041] Specifically, the steps for constructing a faulty spare parts demand fluctuation prediction model based on window grid confidence values, demand occurrence window tables, and inventory flow data are as follows: Using window grid confidence values, demand occurrence window tables, consumption verification cycles, standardized fault point records, and inventory flow data as basic inputs, the cycle numbers of the confirmation event codes and the outbound event codes are associated and paired according to the maintenance order number. A time-series bucketing coding method is used to convert the cycle number difference between the confirmation event code and the outbound event code into a time-delay bucket sequence. The bucket boundaries of the time-series bucketing coding method are determined by the distribution of the cycle number difference between the confirmation event code and the outbound event code in the historical closed-path samples. Multiple time delays are generated according to the quartiles of the cycle number difference and the window span threshold. When the difference in cycle numbers falls into the corresponding time-delay bucket, the corresponding time-delay bucket number is written. Cycle number differences exceeding the window span threshold are written as the over-limit bucket number. The time-delay bucket sequence is arranged according to the time direction from the confirmation event code to the outbound event code. Sequence elements include the time-delay bucket number, the cycle number of the confirmation event code, the cycle number of the outbound event code, and the consumption verification cycle. The sequence length is determined by the maximum cycle number span corresponding to the window span threshold. When the difference in cycle numbers between the confirmation event code and the outbound event code is less than the sequence length, it is padded with an empty bucket marker. When the difference in cycle numbers is greater than the sequence length, the time-delay bucket numbers within the window span threshold coverage area are retained, and the excess is written as the over-limit bucket number. The confirmation waiting segment duration and the available inventory quantity before and after the window are also considered. The difference, window grid confidence value, time lag bucket number, standard fault point name code, fault group identifier code, spare part number code, outbound quantity and available inventory quantity within the consumption verification cycle are written into the feature matrix. The window grid confidence value is used to characterize the confidence level of the demand occurrence window, the fault point standardization record is used to characterize the semantic normalization result of the fault point, and the inventory flow data is used to characterize the changes in outbound quantity and available inventory quantity. When the confirmation event code in the demand occurrence window table forms an outbound event code matching the spare part number within the corresponding consumption verification cycle, the fluctuation label value is written as a valid fluctuation marker, with a value of 1. When the confirmation event code does not form an outbound event code matching the spare part number within the corresponding consumption verification cycle, the fluctuation label value is written as no fluctuation. The fluctuation label value is set to 0. When multiple records of the same maintenance order number, the same fault group identifier, and the same spare part number appear in the same consumption verification cycle, the sample with the earliest confirmation event code and the outbound event code matching is retained as the training sample, and the remaining duplicate samples are written into the deduplication sample record. The gradient boosting decision tree method is used to construct a fault spare part demand fluctuation prediction model based on the feature matrix and fluctuation label value. The window grid reliability analysis results, time lag bucket sequence, fluctuation label value construction, and inventory flow characteristics jointly participate in the fault spare part demand fluctuation prediction and output the spare part demand fluctuation value. The spare part demand fluctuation value is a probability value between 0 and 1, which is used to represent the degree of possibility of spare part demand fluctuation in the corresponding fault group identifier and spare part number in the consumption verification cycle.

[0042] In this implementation scheme, the difference in periodic numbers between the confirmation event code and the outbound event code is transformed into a time-delayed bucket sequence using a time-series bucketing coding method. The window grid reliability value, confirmation waiting segment duration, difference in available inventory quantity before and after the window, standardized fault point records, and inventory flow data are uniformly written into the feature matrix. This allows the variable time-delay relationship between fault confirmation and spare parts outbound to be identified by the faulty spare parts demand fluctuation prediction model. Valid fluctuation markers and non-fluctuation markers are distinguished by fluctuation label values, and duplicate samples under the same maintenance order number, the same fault group identifier, and the same spare parts number are deduplicated, reducing the impact of receiving noise, outbound lag, and duplicate records on model training. This outputs a spare parts demand fluctuation value with a clearly defined range, providing a reliable predictive basis for subsequent fluctuation generation through labeling, safety stock adjustment record writing, and inventory trigger flow diagram construction.

[0043] Specifically, the steps for executing fluctuation tracking and abnormal fluctuation sample recording are as follows: By comparing the spare parts demand fluctuation value and the fluctuation threshold in real time, when the spare parts demand fluctuation value is greater than or equal to the fluctuation threshold, a fluctuation tracking is generated and written to the safety stock adjustment record. The cycle number of the outbound event code is used as the consumption verification cycle, and the corresponding fault group identifier, spare parts number, demand occurrence cycle, consumption verification cycle, spare parts demand fluctuation value, available inventory quantity, and current cycle safety stock threshold are written to the replenishment candidate queue. The replenishment candidate queue uses the fault group identifier and spare parts number as the queue primary key. When the same queue primary key is written repeatedly, the record with the larger spare parts demand fluctuation value is retained, and the records with earlier demand occurrence cycles, available inventory quantities less than the current cycle safety stock threshold, and replenishment arrival cycles further from the consumption verification cycle are arranged at the front of the queue. When the records in the replenishment candidate queue do not show the corresponding outbound event code or fluctuation in subsequent statistical cycles... When a fault is marked as invalid, the record is moved to the candidate expired list to avoid repeated triggering of safety stock adjustments for the same fault group identifier and spare part number. When the spare part demand fluctuation value is less than the fluctuation threshold, a fluctuation pending verification mark is generated, and the safety stock threshold of the previous period is retained. When a spare part number matching the outbound event code occurs in the future, the newly added confirmation event code, outbound event code, time lag bucket sequence, and fluctuation label value are written into the training sample table. After the statistical period ends, the feature matrix is ​​updated based on the training sample table, and the fault spare part demand fluctuation prediction model is incrementally trained and the spare part demand fluctuation value is re-output. If the re-output spare part demand fluctuation value is still less than the fluctuation threshold, the corresponding sample is written into the abnormal fluctuation sample record. The abnormal fluctuation sample record is written with the fault group identifier, spare part number, demand occurrence period, consumption verification period, time lag bucket sequence, and fluctuation label value, and is used for fluctuation threshold verification, training sample screening, and window rearrangement list verification in the next statistical period.

[0044] In this implementation scheme, by comparing the fluctuation value of spare parts demand with the fluctuation threshold, the fault group identifier and spare parts number that meet the fluctuation conditions are written into the replenishment candidate queue. The queue is sorted, deduplicated, and expired based on the demand occurrence cycle, consumption verification cycle, available inventory quantity, and the current period safety stock threshold. This allows the replenishment trigger to prioritize spare parts records with earlier demand, insufficient inventory coverage, and delayed delivery cycles. For samples that do not reach the fluctuation threshold, a second judgment is made through fluctuation verification marking, training sample table, feature matrix update, and incremental training. Samples that still do not reach the fluctuation threshold are written into the abnormal fluctuation sample record for subsequent fluctuation threshold verification, training sample screening, and window rearrangement list verification. This reduces the interference of occasional maintenance records, unfulfilled fault records, and duplicate candidate records on safety stock adjustments.

[0045] Specifically, the steps for constructing an inventory trigger flow diagram based on fluctuation pass markers, spare parts demand fluctuation values, safety stock adjustment records, and inventory flow data are as follows: The inventory trigger flow diagram is constructed based on fluctuation pass markers, spare parts demand fluctuation values, safety stock adjustment records, available inventory quantity, current period safety stock threshold, demand occurrence cycle, consumption verification cycle, replenishment order generation cycle, and replenishment arrival cycle. The inventory trigger flow diagram adopts a directed graph structure. Graph nodes include demand window write nodes, fluctuation pass nodes, inventory deduction nodes, replenishment generation nodes, and arrival backfill nodes. Graph edges are arranged according to the following order: demand window write node to fluctuation pass node, fluctuation pass node to inventory deduction node, and inventory deduction node to replenishment generation node. The business flow sequence from the replenishment generation node to the arrival and backfilling node is connected. Each graph edge is written with the cycle number span, change in available inventory quantity, and spare parts demand fluctuation value as path weights. The corresponding graph edge is refreshed by spare parts number after a new fluctuation pass marker, safety stock adjustment record, or inventory flow data is added. Based on the demand occurrence cycle, records where spare parts demand fluctuation value is greater than or equal to the fluctuation threshold, consumption verification cycle, replenishment order generation cycle, and replenishment arrival cycle, demand window write nodes, fluctuation pass nodes, inventory deduction nodes, replenishment generation nodes, and arrival and backfilling nodes are generated respectively. An inventory level trajectory table is constructed using spare parts number as the row index and cycle number as the column index, and the demand window write node, fluctuation pass node, and arrival and backfilling node are used as the path weights. Three trigger paths, with inventory deduction nodes as replenishment starting points, are replayed in the inventory level trajectory table. The trigger path starting at the demand window write node is used to verify the risk of inventory buildup caused by early demand occurrence. The trigger path starting at the fluctuation pass node is used to verify the alignment of the replenishment trigger after the spare parts demand fluctuation value passes. The trigger path starting at the inventory deduction node is used to verify the risk of spare parts shortage caused by consumption verification cycle triggering. The path corresponding to the fluctuation pass node is used as the default candidate path. When multiple paths generate replenishment trigger records within the same cycle number, the trigger path with the smallest coverage quantity corresponding to the inventory trajectory alignment value is prioritized. Before the replenishment arrival cycle, the demand occurrence cycle to the consumption... When a grid within the verification cycle coverage area has an available inventory quantity less than the current cycle's safety stock threshold, and the corresponding grid experiences fluctuations via a flag or outbound event code, the corresponding grid is written as a shortage grid and a delayed trigger flag is written. When, after the replenishment cycle, the available inventory quantity minus the outbound quantity corresponding to the first outbound event code after the replenishment cycle is still greater than the current cycle's safety stock threshold, the corresponding grid is written as a backlog grid and an early trigger flag is written. When the same spare part number experiences alternating writing of threshold upward adjustment flag, threshold downward adjustment flag, and threshold upward adjustment flag within three adjacent cycles, or alternating writing of threshold downward adjustment flag, threshold upward adjustment flag, and threshold downward adjustment flag, a threshold swing flag is written.The inventory trajectory alignment value is obtained based on the coverage quantity and position of the lagging trigger marker, early trigger marker, and threshold swing marker in the arrangement of the three trigger paths. The inventory trajectory alignment value is obtained by subtracting the coverage quantity deduction item and the key cycle coverage position deduction item from one, with a value range of 0 to 1. When the lagging trigger marker is located in a coverage position before the replenishment arrival cycle, the early trigger marker is located in a coverage position after the replenishment arrival cycle, and the threshold swing marker is located in an alternating position between adjacent cycles, the corresponding deduction item is added, so that the inventory trajectory alignment value can reflect the degree of alignment between the trigger path and the change in inventory level.

[0046] In this implementation plan, a replayable trigger path is created by connecting the demand window writing node, fluctuation passing node, inventory deduction node, replenishment generation node, and arrival backfill node through an inventory trigger flow diagram. These three trigger paths are then mapped to an inventory level trajectory table for comparison and verification, allowing for the quantification of differences between demand occurrence cycle triggers, spare parts demand fluctuation value triggers, and consumption verification cycle triggers. Shortage grids, overstock grids, and threshold swing markers are used to identify replenishment lag, replenishment advances, and repeated changes in the safety stock threshold, respectively. The inventory trajectory alignment value is obtained based on the coverage quantity and position of these three types of markers in the three trigger paths, providing a traceable basis for judgment regarding replenishment trigger start point migration, path rearrangement, and the writing of the current cycle's safety stock threshold. Figure 4 The figure shows a comparison of inventory level trajectories under different replenishment trigger paths in this embodiment. The horizontal axis represents the cycle number, and the vertical axis represents the available inventory quantity. It shows the changes in the available inventory quantity under three trigger paths: the demand window write node, the fluctuation pass node, and the inventory deduction node, with the safety stock threshold as the comparison benchmark. The figure marks the replenishment order generation cycle and the replenishment arrival cycle. The delayed trigger mark indicates the grid where the available inventory quantity is less than the safety stock threshold before the replenishment arrival cycle. The advanced trigger mark indicates the grid where the available inventory quantity is still greater than the safety stock threshold after deducting the outbound quantity after the replenishment arrival cycle. The threshold swing mark indicates the situation where the threshold adjustment mark and the threshold reduction mark are written alternately in adjacent cycles. This shows the differences of the three trigger paths in terms of parts shortage, inventory backlog, and threshold swing, providing a data example for obtaining the inventory trajectory alignment value.

[0047] Specifically, the steps for executing trigger start point migration and path rearrangement are as follows: By comparing the inventory trajectory alignment value and alignment threshold in real time, when the inventory trajectory alignment value is greater than or equal to the alignment threshold, the trigger path with the fewest covered quantities of lagging trigger markers, early trigger markers, and threshold swing back markers is selected; when multiple trigger paths have the same coverage quantity, the path is filtered in the order of fewest threshold swing back markers, fewest lagging trigger markers, coverage position close to the consumption verification cycle, and large spare parts demand fluctuation value, and the selected trigger path is written into the safety stock adjustment record to generate a replenishment order instruction; when the inventory trajectory alignment value is less than the alignment threshold, the lagging trigger markers, early trigger markers, and threshold swing back markers in the arrangement of the three trigger paths are read; the missing parts grid is the grid where the available inventory quantity before the replenishment arrival cycle is less than the safety stock threshold of the current cycle, and the overstock grid is the grid where the available inventory quantity after deducting the outbound quantity after the replenishment arrival cycle is still greater than the safety stock threshold of the current cycle; for trigger paths carrying lagging trigger markers... The replenishment trigger point is moved forward from the inventory deduction node to the fluctuation passage node. If there are still missing parts grids in the corresponding grids of the fluctuation passage node, it is moved forward to the demand window write node, and replenishment trigger records are generated according to the demand occurrence cycle. For trigger paths with an early trigger mark, the replenishment trigger point is moved backward from the demand window write node to the fluctuation passage node. If there are still overstock grids after the replenishment arrival cycle, it is moved backward to the inventory deduction node, and replenishment trigger records are generated according to the consumption verification cycle. For trigger paths with a threshold swing-back mark, the current cycle's safety stock threshold writing is stopped, the previous cycle's safety stock threshold is used, and the corresponding fault group identifier, spare part number, replenishment trigger point, and mark type are written into the safety stock adjustment record. Records that use the previous cycle's safety stock threshold are only effective in the current cycle. The inventory trajectory fit value is reread in the next cycle. When the next cycle's inventory trajectory fit value is greater than or equal to the fit threshold, and the corresponding trigger path does not carry a threshold swing-back mark, the current cycle's safety stock threshold writing is resumed.

[0048] In this implementation plan, the three trigger paths are screened and rearranged by comparing the inventory trajectory alignment value with the alignment threshold. This prioritizes trigger paths with fewer covered items (lagging trigger markers, early trigger markers, and threshold reversal markers) for safety stock adjustments. When paths are parallel, the replenishment trigger path is further determined based on marker type, coverage location, and spare parts demand fluctuation value to avoid inconsistent replenishment trigger starting point selection. When a trigger path has a lagging trigger marker, the replenishment trigger starting point is moved forward to reduce the risk of replenishment lag. When a trigger path has an early trigger marker, the replenishment trigger starting point is moved backward to reduce the risk of overstocking. When a trigger path has a threshold reversal marker, the safety stock threshold of the previous period is used and reassessed in the next period, so that the safety stock threshold adjustment is constrained by the inventory trajectory, thereby reducing the impact of early replenishment, late replenishment, and repeated threshold changes on the dynamic adjustment of spare parts safety stock.

[0049] Specifically, this embodiment provides a spare parts safety stock dynamic adjustment system based on demand fluctuation forecasting, applied to a spare parts safety stock dynamic adjustment method based on demand fluctuation forecasting. It includes a maintenance timeline acquisition and preprocessing module, used to collect hardware maintenance demand data, inventory flow data, fault point terminology data, and historical safety stock records, and to perform standardization and normalization processing. In a computer hardware maintenance scenario, it obtains the maintenance order number, equipment brand, equipment model, equipment batch, original fault name, arrival time, initial inspection time, fault confirmation time, and maintenance completion time through maintenance work orders. It obtains spare part number, spare part name, outbound time, outbound quantity, available inventory quantity, current period safety stock threshold, replenishment order generation time, and replenishment arrival time through inventory flow data. It completes a unified mapping of standard fault point names, synonyms, prohibited terms, and fault point-related spare part numbers through fault point thesaurus data. It reads the previous period's safety stock threshold and replenishment trigger point from historical safety stock records, enabling hardware fault records from different brands, models, and batches to be organized according to a unified period number, fault group identifier, and fault spare part index table; demand window grid. The reliability analysis module performs demand window grid reliability analysis based on the matching relationship and path length between confirmation event codes and outbound event codes. Based on the grid reliability analysis results, it performs demand occurrence cycle confirmation and replenishment candidate blocking. It records arrival event codes, initial inspection event codes, confirmation event codes, outbound event codes, and completion event codes through a window path matrix, identifying candidate closed paths, drift span markers, and noise / hole markers between confirmation event codes and outbound event codes. This ensures that unconfirmed faults in the arrival phase and delayed deduction records in the outbound phase do not directly interfere with the demand occurrence cycle. A faulty spare parts demand fluctuation prediction model is also included. The module is used to build a prediction model for the fluctuation of spare parts demand based on the window grid confidence value, the demand occurrence window table and inventory flow data. The fluctuation is executed by writing the mark and recording the abnormal fluctuation sample. The difference of the period number between the confirmation event code and the outbound event code is converted into a time-delay bucket sequence through the time-series bucket coding method. The confirmation waiting segment duration, the difference of the available inventory quantity before and after the window, the standardized record of the fault point and the inventory flow data are written into the feature matrix, so that the spare parts demand fluctuation value can reflect the degree of possibility of the fault grouping identifier and spare parts number forming a change in spare parts demand within the consumption verification cycle.The safety stock trigger trajectory correction module is used to construct an inventory trigger flow diagram based on fluctuation passing markers, spare parts demand fluctuation values, safety stock adjustment records, and inventory flow data. It performs trigger start point migration and path rearrangement, constructing a trigger path through demand window write nodes, fluctuation passing nodes, inventory deduction nodes, replenishment generation nodes, and arrival backfill nodes. It identifies lagging trigger markers, early trigger markers, and threshold reversal markers in the inventory level trajectory table, enabling the replenishment trigger start point to migrate between demand window write nodes, fluctuation passing nodes, and inventory deduction nodes. This ensures that the current cycle's safety stock threshold and replenishment order generation cycle are consistent with the hardware maintenance spare parts consumption rhythm.

[0050] This implementation plan establishes a complete processing chain from data acquisition, window discrimination, fluctuation prediction to inventory trigger correction through four modules: a maintenance timeline acquisition and preprocessing module, a demand window grid reliability analysis module, a faulty spare parts demand fluctuation prediction model construction module, and a safety stock trigger trajectory correction module. This enables hardware maintenance demand data, inventory flow data, fault point terminology data, and historical safety stock records to establish a unified association around fault grouping identifiers, spare parts numbers, and cycle numbers. The demand window grid reliability analysis eliminates interference from unconfirmed faults during the receiving stage and delayed deductions during the outbound stage on the demand occurrence cycle. The faulty spare parts demand fluctuation prediction model outputs spare parts demand fluctuation values, providing a basis for judgment on fluctuation passing marks and abnormal fluctuation sample records. Finally, the inventory trigger flow diagram corrects the trigger paths corresponding to the demand window write nodes, fluctuation passing nodes, and inventory deduction nodes, ensuring that the replenishment trigger start point, the current cycle safety stock threshold, and the replenishment order generation cycle match the actual consumption rhythm of computer hardware maintenance spare parts, reducing inventory adjustment deviations caused by replenishment lag, premature inventory buildup, and threshold reversals.

[0051] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0052] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for dynamically adjusting spare parts safety stock based on demand fluctuation forecasting, characterized in that, Includes the following steps: S1 collects hardware repair demand data, inventory flow data, fault point terminology data, and historical safety stock records, and performs standardization and normalization processing. S2, based on the matching relationship between the confirmation event code and the outbound event code and the path length, perform demand window grid reliability analysis, and execute demand occurrence cycle confirmation and replenishment candidate blocking based on the grid reliability analysis results; S3 builds a faulty spare parts demand fluctuation prediction model based on window grid confidence value, demand occurrence window table and inventory flow data, and executes fluctuations by writing tags and recording abnormal fluctuation samples. S4 constructs an inventory trigger flow graph based on fluctuation markers, spare parts demand fluctuation values, safety stock adjustment records, and inventory flow data, and performs trigger start point migration and path rearrangement.

2. The method for dynamically adjusting spare parts safety stock based on demand fluctuation forecasting according to claim 1, characterized in that: The specific steps for collecting hardware repair demand data, inventory flow data, fault point terminology data, and historical safety stock records are as follows: Collect hardware repair demand data: Read the repair order number, equipment brand, equipment model, equipment batch, original name of the fault point, arrival time, initial inspection time, fault confirmation time and repair completion time from the repair work order; Collect inventory flow data: Read spare parts number, spare parts name, outbound time, outbound quantity, available inventory quantity, current period safety stock threshold, replenishment order generation time, and replenishment arrival time from the inventory flow data; Collect fault point terminology data: Read standard fault point names, synonym entries, prohibited entries, and fault point associated spare part numbers from the fault point terminology; Collect historical safety stock records: Read spare parts number, previous cycle safety stock threshold and replenishment trigger point from historical safety stock records.

3. The method for dynamically adjusting spare parts safety stock based on demand fluctuation forecasting according to claim 1, characterized in that: The specific steps for standardization and normalization are as follows: The hardware repair demand data, inventory flow data, fault point thesaurus data, and historical safety stock records are cleaned, deduplicated, time-standardized, and field-normalized to generate a unified cycle number; a replenishment generation mark is generated based on the replenishment order generation time, and an arrival backfill mark is generated based on the replenishment arrival time; the original fault point names are mapped to standard fault point names to generate standardized fault point records. Generate fault grouping identifiers based on equipment brand, equipment model, equipment batch, and standard fault point name; Write records where the current period's safety stock threshold is greater than the previous period's safety stock threshold to the threshold increase flag, and write records where the current period's safety stock threshold is less than the previous period's safety stock threshold to the threshold decrease flag. A time chain record is generated based on the repair order number, cycle number, receiving time, initial inspection time, fault confirmation time, outbound time, and repair completion time, and receiving event code, initial inspection event code, confirmation event code, outbound event code, and completion event code are written in. A confirmation waiting segment duration is generated based on the time interval between the fault confirmation time and the outbound time. Candidate fields for demand occurrence cycle and consumption verification cycle are generated based on the confirmation event code and the outbound event code. A replenishment order generation cycle is generated based on the replenishment order generation time, and a replenishment arrival cycle is generated based on the replenishment arrival time. The difference in available inventory quantity before and after the window is generated based on the available inventory quantity of adjacent cycles. A faulty spare parts index table is generated based on the fault group identifier, standard fault point name, fault point associated spare parts number, and spare parts number.

4. The method for dynamically adjusting spare parts safety stock based on demand fluctuation forecasting according to claim 1, characterized in that: The specific steps for performing demand window grid reliability analysis based on the matching relationship between confirmation event codes and outbound event codes and path length are as follows: A window path matrix is ​​constructed using the fault group identifier and maintenance order number as a combined row index and the cycle number as a column index; the receiving event code, initial inspection event code, confirmation event code, outbound event code, and completion event code are written into the window path matrix; Read the path length between the confirmation event code and the outbound event code along the same maintenance order number, and verify the matching relationship between the standard fault point name corresponding to the confirmation event code and the spare part number in the fault spare part index table; when the confirmation event code is before the outbound event code and the spare part number matching relationship is valid, write the grid segment between the confirmation event code and the outbound event code into the candidate closed path; when the path length of the candidate closed path does not exceed the window span threshold, write the candidate closed path into the closed path marker and write the corresponding window into the trusted window queue; when the path length of the candidate closed path exceeds the window span threshold, write the candidate closed path into the drift span marker and write the corresponding window into the drift window queue; when there is no confirmation event code between the receiving event code and the outbound event code, write the corresponding grid segment into the noise hole marker and write the corresponding window into the noise window queue. Based on the coverage positions of closed path markers, noise hole markers, and drift span markers under the same fault group identifier, the window grid confidence value is obtained.

5. The method for dynamically adjusting spare parts safety stock based on demand fluctuation forecasting according to claim 1, characterized in that: The specific steps for confirming the demand occurrence cycle and blocking replenishment candidates based on the grid reliability analysis results are as follows: By comparing the window grid confidence value and confidence threshold in real time, when the window grid confidence value is greater than or equal to the confidence threshold, the corresponding grid segment is written into the drift window queue, the cycle number where the confirmation event code is located is determined as the demand occurrence cycle, and the corresponding fault group identifier and spare part number are input into the fault spare part demand fluctuation prediction model. When the window grid confidence value is less than the confidence threshold, traffic splitting is performed according to the path marker type; If the drift span marker covers the grid segment between the confirmation event code and the outbound event code, create a demand occurrence window table with the cycle number where the confirmation event code is located, and write the cycle number where the outbound event code is located as the consumption verification cycle. If a closed path marker is missing, the safety stock threshold from the previous period is retained, and the corresponding grid segment is written to the window rearrangement list.

6. The method for dynamically adjusting spare parts safety stock based on demand fluctuation forecasting according to claim 1, characterized in that: The specific steps for constructing a faulty spare parts demand fluctuation prediction model based on window grid confidence values, demand occurrence window tables, and inventory flow data are as follows: Using window grid confidence value, demand occurrence window table, consumption verification cycle, fault point standardized record and inventory flow data as basic inputs, the cycle number of the confirmation event code and the cycle number of the outbound event code are associated and paired according to the maintenance order number. The time-series bucket coding method is used to convert the cycle number difference between the confirmation event code and the outbound event code into a time-delay bucket sequence. The confirmation waiting segment duration and the difference in available inventory quantity before and after the window are written into the feature matrix. When the confirmation event code in the demand occurrence window table generates an outbound event code that matches the spare parts number within the corresponding consumption verification cycle, the fluctuation tag value will be written as a valid fluctuation marker. When the confirmation event code does not generate an outbound event code matching the spare part number within the corresponding consumption verification cycle, the fluctuation tag value is written as a non-fluctuation marker; the gradient boosting decision tree method is used to construct a fault spare part demand fluctuation prediction model based on the feature matrix and fluctuation tag value, and output the spare part demand fluctuation value.

7. The method for dynamically adjusting spare parts safety stock based on demand fluctuation forecasting according to claim 1, characterized in that: The specific steps for recording execution fluctuations through marker writing and abnormal fluctuation sample recording are as follows: By comparing the fluctuation value of spare parts demand with the fluctuation threshold in real time, when the fluctuation value of spare parts demand is greater than or equal to the fluctuation threshold, a fluctuation pass is generated and written into the safety stock adjustment record, and the cycle number of the outbound event code is used as the consumption verification cycle. When the fluctuation value of spare parts demand is less than the fluctuation threshold, a fluctuation pending verification mark is generated, and the safety stock threshold of the previous period is retained. When a spare part number matching the outbound event code appears in the future, the newly added confirmation event code, outbound event code, time delay bucket sequence and fluctuation label value will be written back to the faulty spare part demand fluctuation prediction model. If the re-output spare part demand fluctuation value is still less than the fluctuation threshold, the corresponding sample will be written into the abnormal fluctuation sample record.

8. The method for dynamically adjusting spare parts safety stock based on demand fluctuation forecasting according to claim 1, characterized in that: The specific steps for constructing an inventory trigger flow diagram based on fluctuations, markers, spare parts demand fluctuation values, safety stock adjustment records, and inventory flow data are as follows: An inventory trigger flow diagram is constructed based on fluctuation pass markers, spare parts demand fluctuation values, safety stock adjustment records, available inventory quantity, current period safety stock threshold, demand occurrence cycle, consumption verification cycle, replenishment order generation cycle, and replenishment arrival cycle. Based on the demand occurrence cycle, records where spare parts demand fluctuation values ​​are greater than or equal to the fluctuation threshold, consumption verification cycle, replenishment order generation cycle, and replenishment arrival cycle, demand window write nodes, fluctuation pass nodes, inventory deduction nodes, replenishment generation nodes, and replenishment return nodes are generated respectively. An inventory level trajectory table is constructed using spare parts number as the row index and cycle number as the column index, and the three trigger paths starting from the demand window write node, fluctuation pass node, and inventory deduction node are replayed to the inventory level trajectory table. When a grid appears before the replenishment arrival cycle where the available inventory quantity is less than the current period safety stock threshold, a delayed trigger marker is written. When a grid appears after the replenishment cycle where the available inventory quantity minus the outbound quantity is still greater than the safety stock threshold for the current cycle, an early trigger flag is written; when threshold adjustment flags and threshold reduction flags are written alternately in adjacent cycles, a threshold swing-back flag is written. Based on the coverage quantity and coverage position of the delayed trigger marker, the early trigger marker, and the threshold swing marker in the arrangement of the three trigger paths, the inventory trajectory fit value is obtained.

9. The method for dynamically adjusting spare parts safety stock based on demand fluctuation forecasting according to claim 8, characterized in that: The specific steps for the execution trigger starting point migration and path rearrangement are as follows: By comparing the inventory trajectory alignment value and the alignment threshold in real time, when the inventory trajectory alignment value is greater than or equal to the alignment threshold, the trigger path with the fewest covered quantities of the delayed trigger mark, the early trigger mark, and the threshold swing mark is selected, and the selected trigger path is written into the safety stock adjustment record to generate a replenishment order instruction. When the inventory trajectory alignment value is less than the alignment threshold, read the delayed trigger flag, the early trigger flag, and the threshold swing flag from the three trigger paths. For trigger paths with a delayed trigger flag, the replenishment trigger start point is moved forward from the inventory deduction node to the fluctuation passing node. If there are still missing parts grids in the corresponding grids of the fluctuation passing node, it is moved forward to the demand window write node, and a replenishment trigger record is generated according to the demand occurrence cycle. For trigger paths with an early trigger flag, the replenishment trigger start point is moved backward from the demand window write node to the fluctuation passing node. If there are still overstock grids after the replenishment arrival cycle, it is moved backward to the inventory deduction node, and a replenishment trigger record is generated according to the consumption verification cycle. For trigger paths carrying threshold swing-back markers, stop writing the current period's safety stock threshold, use the previous period's safety stock threshold, and write the corresponding fault group identifier, spare part number, replenishment trigger start point, and marker type into the safety stock adjustment record.

10. A spare parts safety stock dynamic adjustment system based on demand fluctuation forecasting, applied to the spare parts safety stock dynamic adjustment method based on demand fluctuation forecasting as described in any one of claims 1-9, characterized in that, include: The maintenance timeline acquisition and preprocessing module is used to collect hardware maintenance demand data, inventory flow data, fault point terminology data and historical safety stock records, and to perform standardization and normalization processing. The demand window grid reliability analysis module is used to perform demand window grid reliability analysis based on the matching relationship between confirmation event codes and outbound event codes and the path length, and to perform demand occurrence cycle confirmation and replenishment candidate blocking based on the grid reliability analysis results. The module for building a demand fluctuation prediction model for faulty spare parts is used to build a demand fluctuation prediction model for faulty spare parts based on the window grid confidence value, the demand occurrence window table and inventory flow data. The fluctuation is executed by writing the flag and recording the abnormal fluctuation samples. The safety stock trigger trajectory correction module is used to construct an inventory trigger flow diagram based on fluctuation pass markers, spare parts demand fluctuation values, safety stock adjustment records, and inventory flow data, and to perform trigger start point migration and path rearrangement.

Images