A mold storage management method and system based on big data analysis

Abstract

The application provides a mold warehouse management method based on big data analysis, which collects mold full life cycle heterogeneous data through a multi-source collection terminal, obtains a standardized data set through standardized cleaning and correlation mapping, calculates mold health grades and residual service life through a health degree dynamic attenuation evaluation model, solves a globally optimal storage location allocation scheme through a multi-factor weighted adaptive algorithm, generates an optimal access execution sequence through a multi-constraint priority scheduling mechanism, and collects state data in real time and calculates deviations during the execution process to generate dynamic correction instructions. The application can realize mold warehouse full link intelligent closed loop management and control, and improves warehouse space utilization, access efficiency and mold full life cycle management accuracy.

Description

Technical Field

[0001] This invention relates to the field of mold storage management technology, and in particular to a mold storage management method and system based on big data analysis. Background Technology

[0002] As a core process equipment in the equipment manufacturing industry, molds are characterized by their wide variety, large size differences, high unit value, and the need for complete sets of equipment. The efficiency and precision of their warehousing management directly affect production cycle time, product quality, and enterprise manufacturing costs. With the advancement of intelligent manufacturing technology, mold warehousing management is gradually upgrading from manual ledgers and barcode management to digitalization and intelligence.

[0003] Existing mold storage management solutions are mostly based on the transformation of general storage systems. They use anti-metal RFID technology to identify molds and register their entry and exit from the warehouse, but can only complete basic inventory ledger management and static storage location allocation. Such solutions have several significant shortcomings: 1) Insufficient depth of big data analysis, only basic data statistics such as usage frequency and inventory quantity can be achieved, but it cannot integrate multi-dimensional data such as production wear and tear, environmental erosion, and maintenance and repair to accurately assess the health status of molds. Life prediction and maintenance management are highly dependent on human experience, which can easily lead to the dual problems of over-maintenance increasing costs or untimely maintenance leading to defective products; 2) Fixed classification rules are used for warehouse location allocation, which do not fully adapt to the unique characteristics of irregular shapes, heavy loads, and complete sets of molds. It is impossible to dynamically optimize the warehouse layout by combining storage and retrieval frequency and health level, resulting in generally low warehouse space utilization and storage and retrieval efficiency; 3) Data is disconnected from the warehouse system and the production MES, maintenance management, and quality inspection systems, forming obvious data silos. The linkage between warehouse scheduling and production scheduling is poor, and the problems of discrepancies between accounts and actual inventory and delayed response to anomalies are prominent; 4) Lack of a closed-loop feedback and correction mechanism in the execution process, insufficient robustness of system control, and difficulty in meeting the refined and intelligent control requirements of mold warehousing in modern intelligent manufacturing scenarios. Summary of the Invention

[0004] The purpose of this invention is to provide a mold storage management method and system based on big data analysis, so as to improve the utilization rate of storage space, storage and retrieval efficiency and the accuracy of mold life cycle management.

[0005] To achieve the above objectives, the present invention provides the following solution: A mold storage management method based on big data analysis includes the following steps: Multi-source heterogeneous data is collected from multiple sources throughout the entire life cycle of the mold using a multi-source acquisition terminal, and then standardized cleaning and association mapping are performed to obtain a standardized dataset. The standardized dataset was subjected to multi-factor decay calculations using a dynamic decay assessment model for mold health, resulting in the health level label and predicted remaining service life for each mold. The optimal storage location allocation scheme is obtained by calculating the storage location fit degree based on health level labels, remaining useful life prediction values, storage space topology data and historical access trajectory data through a multi-factor weighted adaptation algorithm. By using a multi-constraint priority scheduling mechanism, the production scheduling task data, the optimal warehouse location allocation scheme, and the health level label are sorted and the path is planned to obtain the optimal storage and retrieval execution sequence. Based on the optimal access execution sequence, access operations are performed through warehouse execution equipment, while execution status data is collected in real time and deviation is calculated to obtain dynamic correction instructions.

[0006] Optionally, the multi-source data acquisition terminals include: RFID readers, environmental sensors, production MES systems, maintenance management systems, and quality inspection systems; the multi-source heterogeneous data includes: mold basic attribute data, warehousing environment data, production and usage data, maintenance record data, and product quality data.

[0007] Optionally, multi-source heterogeneous data is collected throughout the entire lifecycle of the mold using a multi-source acquisition terminal, and then standardized, cleaned, and correlated to obtain a standardized dataset, including: The missing points of continuous time series data in multi-source heterogeneous data are fitted and filled by Lagrange interpolation, and the missing items of discrete attribute data in multi-source heterogeneous data are filled with default values ​​by rule matching, thus obtaining the complete dataset. Outlier removal was performed on the completed dataset using the 3σ criterion to obtain a cleaned dataset; The normalized dataset is obtained by linearly transforming all numerical data of different dimensions in the cleaned dataset using the max-min normalization method. Using the unique RFID code of each mold as the primary key identifier, the associated identifier fields of the mold's basic attributes, component list, production and usage records, maintenance and repair records, and product quality data in the normalized dataset are extracted respectively. Based on the associated identifier field, a hierarchical mapping relationship is constructed to obtain the data mapping chain; the data mapping chain has a five-level hierarchical structure, with the following order: mold body layer, component layer, usage record layer, maintenance record layer, and quality association layer; By binding and storing the data mapping chain with the normalized dataset, a standardized dataset is obtained.

[0008] Optionally, the expression for the dynamic decay assessment model of mold health is: ;in, This represents the current real-time health status of the mold. This is the initial health baseline value after the mold has passed factory acceptance. The cumulative effective number of stampings for the mold. The weighted value of the cumulative wear of the mold. This represents the cumulative value of erosion in the storage environment. This represents the cumulative gain from maintenance and repair. , , and All are weighting coefficients.

[0009] Optionally, multi-factor decay calculations are performed on the standardized dataset to obtain the health level label and remaining service life prediction value for each mold, including: The real-time health value is matched with a preset threshold range to obtain a health level label; the health level labels include: excellent, good, warning, and obsolescence. Based on historical production data in a standardized dataset, the average rate of health decay of the mold per unit time is calculated. Combined with real-time health values ​​and scrap thresholds, the remaining number of stamping cycles and remaining usable time of the mold are estimated to obtain the predicted value of remaining service life.

[0010] Optionally, a multi-factor weighted adaptation algorithm is used to calculate the storage location fit based on health level labels, remaining useful life predictions, storage space topology data, and historical access trajectory data to obtain the optimal storage location allocation scheme, including: The location fit is calculated using a multi-factor weighted fit algorithm; the formula for calculating the location fit is: ;in, To access the frequency adaptation factor, As a health level adaptation factor, As a spatial carrying capacity adaptation factor, For a complete set of related adaptation factors, , , and All are weighting coefficients; The objective function is to maximize the sum of the storage location adaptability of all mold and storage location combinations, and a preset constraint condition is added to the objective function to construct an integer programming solution model. The branch and bound method is used to solve the integer programming model through a traversal process to obtain the optimal allocation combination; When the health level label changes, the production plan is adjusted, or the storage location fails, a real-time recalculation is immediately triggered to update the allocation scheme and obtain the optimal storage location allocation scheme.

[0011] Optionally, the calculation process for the complete set of related adaptation factors includes: The complete set identification field identifies the complete set of molds to which the current mold belongs, and obtains the allocated storage location information of all mold components in the same set; Calculate the average path distance between the current candidate storage location and all allocated component storage locations in the same group based on the allocated storage location information, and determine the set association adaptation factor based on the average path distance.

[0012] Optionally, a multi-constraint priority scheduling mechanism is used to sort and plan the production scheduling task data, the optimal warehouse location allocation scheme, and the health level labels to obtain the optimal access execution sequence, including: The priority value of a single access task is calculated using the following formula: ;in, As a factor of production task urgency, For quality requirement adaptation factors, This is the path efficiency factor. , and All are weighting coefficients; All pending access tasks are initially sorted from high to low priority values ​​to obtain an initial task sequence. The genetic algorithm is used to globally optimize the transport path corresponding to the initial task sequence, merge tasks in the same direction, avoid congested road segments, and minimize the total transport path length to obtain the optimized task sequence. When an emergency production order is received, the priority values ​​of all pending tasks and the newly added emergency task are recalculated, the sorting results are updated, and the subsequent task sequence is adjusted to obtain the optimal access and execution sequence.

[0013] Optionally, based on the optimal access execution sequence, access operations are performed through warehouse execution equipment, while execution status data is collected in real time and deviations are calculated to obtain dynamic correction instructions, including: The optimal access execution sequence is broken down into single-step equipment action instructions and sent to AGV handling equipment and automated warehouse stacker crane to perform access operations; The execution status data is obtained by collecting position data, weight data and attitude data during the execution process through a UWB positioning unit, a weight sensor and a vision recognition unit, respectively. The position deviation, duration deviation, and attitude deviation are calculated separately, and the overall execution deviation is obtained by weighted summation. When the overall execution deviation is less than the preset first threshold, only the deviation data is recorded and no correction instruction is generated; when the overall execution deviation is between the first threshold and the preset second threshold, a real-time correction instruction is generated on the device side to adjust the handling speed and pick-up / placement position; when the overall execution deviation is greater than the second threshold, the device operation is immediately suspended and a manual intervention warning is triggered.

[0014] A mold storage management system based on big data analytics includes: The data preprocessing module is used to collect multi-source heterogeneous data throughout the entire life cycle of the mold through a multi-source acquisition terminal, and to perform standardized cleaning and association mapping to obtain a standardized dataset. The multi-factor calculation module is used to perform multi-factor decay calculations on the standardized dataset through the dynamic decay assessment model of mold health, and obtain the health level label and remaining service life prediction value of each mold. The allocation scheme calculation module is used to calculate the storage location fit degree based on health level labels, remaining useful life prediction values, storage space topology data and historical access trajectory data through a multi-factor weighted adaptation algorithm, so as to obtain the optimal storage location allocation scheme. The sequence optimization module is used to sort and plan the tasks based on the production scheduling task data, the optimal warehouse location allocation scheme and the health level label through a multi-constraint priority scheduling mechanism, so as to obtain the optimal storage and retrieval execution sequence. The feedback correction module is used to perform storage and retrieval operations through the warehouse execution equipment based on the optimal storage and retrieval execution sequence, while simultaneously collecting execution status data in real time and calculating deviations to obtain dynamic correction instructions.

[0015] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects: The present invention breaks down data silos in warehousing, production, maintenance, and quality systems through the full-link fusion of multi-source heterogeneous data, realizing traceability of the entire life cycle of molds; based on a multi-factor dynamic decay model, it accurately assesses the health and remaining life of molds, promoting the upgrade of maintenance mode from fixed cycle to predictive, reducing the cost of excessive maintenance and the rate of defective products; at the same time, it integrates dynamic storage location allocation and multi-constraint scheduling based on dimensions such as health level, access frequency, and complete set association, significantly improving the utilization rate of storage space and access efficiency; coupled with an execution deviation closed-loop feedback correction mechanism, it enhances the robustness and anomaly response capability of the system, realizing refined and intelligent management and control of mold warehousing. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of a mold storage management method based on big data analysis, as described in an embodiment of the present invention. Figure 2 This is a schematic diagram of the mold storage management system based on big data analysis, as described in an embodiment of the present invention. Detailed Implementation

[0018] 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.

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] like Figure 1 As shown in the figure, this invention provides a mold storage management method based on big data analysis, including the following steps: Step 100: Collect multi-source heterogeneous data throughout the entire lifecycle of the mold using a multi-source acquisition terminal, and perform standardized cleaning and association mapping to obtain a standardized dataset; Step 200: Perform multi-factor decay calculation on the standardized dataset using the dynamic decay assessment model for mold health to obtain the health level label and predicted remaining service life value for each mold. Step 300: Calculate the storage location fit degree using a multi-factor weighted adaptation algorithm on health level labels, remaining useful life prediction values, storage space topology data, and historical storage and retrieval trajectory data to obtain the optimal storage location allocation scheme. Step 400: Use a multi-constraint priority scheduling mechanism to sort and plan the production scheduling task data, the optimal warehouse location allocation scheme, and the health level labels to obtain the optimal storage and retrieval execution sequence; Step 500: Based on the optimal access execution sequence, access operations are performed through the warehouse execution equipment, and the execution status data is collected in real time and the deviation is calculated to obtain dynamic correction instructions.

[0021] In the specific implementation process, step 100 first completes the terminal acquisition and system access of multi-source heterogeneous data. Anti-metal RFID readers are deployed at warehouse entrances and exits, stamping production stations, and maintenance workstations. Each mold is implanted with a uniquely coded anti-metal RFID tag, using the format "category prefix-year of production-serial number". The RFID readers automatically collect basic attribute data of the mold, including mold number, specifications, unit weight, dimensions, manufacturing date, initial accuracy level, and set group identification. Temperature and humidity sensors, dust concentration sensors, and corrosive gas detection sensors are deployed in each storage location of the automated warehouse to continuously collect warehousing environment data. Standardized API interfaces are used to connect to the production MES system, maintenance management system, and quality inspection system to pull production usage data, maintenance record data, and product quality data in real time. Production usage data includes stamping batch, single stamping tonnage, number of stampings per batch, and equipment runtime. Maintenance record data includes maintenance time, maintenance type, replacement parts list, and post-maintenance accuracy test values. Product quality data includes the corresponding batch product dimensional deviation, surface defect rate, and number of defective products. Finally, a multi-source heterogeneous raw dataset covering the entire life cycle of the mold is formed.

[0022] For continuous time-series data in the dataset, including warehouse temperature and humidity time-series data, stamping pressure time-series data, and equipment runtime data, Lagrange interpolation is used to fit and complete missing points. Specifically, missing time points in the time series are first identified, and three valid sampling points before and after the missing point are selected as interpolation nodes. A Lagrange interpolation polynomial of the corresponding order is constructed, and the missing time point is substituted into the polynomial to calculate the fitted and completed value. For discrete attribute data, including fields such as component suppliers, maintenance personnel, and product batch remarks, a rule-matching method is used to fill default values. Based on the mold model, a common component supplier of the same mold model is matched as the default value for missing items. Based on the shift to which the maintenance work order belongs, the corresponding shift personnel information is matched to fill the maintenance personnel field, ensuring that all discrete attribute fields are complete and usable.

[0023] Next, the 3σ criterion is used to automate outlier detection and removal. For each numerical indicator in the dataset, the mean μ and standard deviation σ of all valid samples are calculated to determine the confidence interval for normal data as [μ-3σ, μ+3σ]. All data points are iterated through, and data falling outside the confidence interval are identified as outliers and directly removed. For warehouse humidity data, if a sampling point shows a jump-like extreme value exceeding the 3σ interval, it is identified as data from a momentary sensor malfunction and removed.

[0024] Meanwhile, the maximum-minimum normalization method is used to perform linear transformation on all numerical data of different dimensions and magnitudes, mapping all values ​​to the interval [0,1]. The calculation logic is as follows: for any numerical index, take the minimum and maximum values ​​of all samples of the index, and calculate the normalized value for each original value.

[0025] Furthermore, a five-level hierarchical data mapping chain is constructed based on the normalized dataset, ultimately binding to obtain a standardized dataset. Using the unique RFID code of each mold as the global primary key identifier, the association identifier fields for five types of data are first extracted from the normalized dataset: RFID code, mold model, and set group from the mold's basic attributes are extracted as the core identifier for the mold body layer; component number, mold RFID code, and component model from the component list are extracted as the association identifier for the component layer; production batch number, corresponding mold RFID code, and stamping station number from the production usage record are extracted as the association identifier for the usage record layer; maintenance work order number, corresponding mold RFID code, and replacement component number from the maintenance record are extracted as the association identifier for the maintenance record layer; and quality inspection batch number, corresponding production batch number, and corresponding mold RFID code from the product quality data are extracted as the association identifier for the quality association layer. Subsequently, following the hierarchical order of mold body layer, component layer, usage record layer, maintenance record layer, and quality association layer, a downward mapping relationship is constructed layer by layer: the mold body layer is associated with all components of the mold through RFID coding in a one-to-many manner; the component layer is associated with the corresponding maintenance and replacement records through component numbers; the usage record layer is associated with the corresponding product quality data in a one-to-one manner through production batch numbers; and both the maintenance record layer and the quality association layer can be back-aggregated to the mold body through RFID coding. This ultimately forms a five-level chain-like data mapping structure that is traceable from top to bottom and statistically verifiable from bottom to top. This data mapping chain is then structured and bound to a normalized dataset for storage, thereby obtaining a complete standardized dataset.

[0026] It should be noted that step 100, through multi-terminal, cross-system, multi-dimensional data collection, achieves complete coverage of heterogeneous data throughout the entire lifecycle of molds, from warehousing, use, maintenance to quality feedback. Through a three-level standardized cleaning process combining continuous time-series interpolation and discrete rule filling for differentiated missing value processing, 3σ criterion-based automated anomaly removal, and maximum / minimum dimension normalization, it solves the problems of poor data integrity, high distortion rate, and inconsistent dimensions from multiple sources. Simultaneously, the constructed five-level hierarchical data mapping chain breaks down data silos across the four major systems of warehousing, production, maintenance, and quality, achieving structured association and traceable management of mold data across the entire chain, significantly improving the data granularity and calculation accuracy of mold warehousing control.

[0027] In the specific implementation process, step 200 completes the quantitative assessment of health status and life prediction based on the dynamic decay assessment model of mold health, firstly by initializing the model baseline. Specifically, from the five-level data mapping chain of the standardized dataset, the full life cycle correlation data corresponding to a single mold is extracted, and the initial health baseline value after the mold passes factory acceptance is calibrated. A score of 100 is required. The acceptance criteria cover four core indicators: dimensional accuracy, surface hardness, cavity roughness, and guide clearance. Only after all four indicators meet the standards can the initial baseline value be used. (Model weighting coefficients) , , , The weighting method is determined by combining entropy weighting and expert scoring: First, the entropy weighting method is used to calculate the objective weight of each influencing factor based on the historical full life cycle data of the same batch of molds; then, the weights are weighted and integrated using the experience scores of five senior engineers in the field of stamping die engineering, and finally, the weight coefficients suitable for the specific scenario are determined. For stamping fatigue attenuation weight, For wear and tear attenuation weight, For environmental erosion attenuation weight, Weighting for maintenance and repair gains.

[0028] After completing the baseline initialization, the attenuation factor and gain factor are calculated sequentially using quantization. First, the cumulative effective number of stampings is calculated. Weighted by cumulative wear To address the cumulative effective stamping count, a tonnage-based equivalent conversion mechanism is introduced for loss correction: Using a single stamping operation at the mold's rated stamping tonnage as the baseline equivalent unit, the actual stamping tonnage is graded and converted according to production batches. When the actual stamping tonnage is ≥ 1.1 times the rated tonnage, a single stamping operation is equivalent to 1.5 effective stamping operations; when 0.9 times ≤ actual stamping tonnage < 1.1 times the rated tonnage, a single stamping operation is equivalent to 1 effective stamping operation; when the actual stamping tonnage < 0.9 times the rated tonnage, a single stamping operation is equivalent to 0.7 effective stamping operations. The cumulative effective stamping count is obtained by summing these values ​​batch by batch. This more accurately reflects the differences in the impact of stamping fatigue on mold health under different loads. For the weighted value of cumulative wear, historical dimensional inspection data, cutting edge grinding records, and cavity repair data of the mold are extracted from the standardized dataset. Wear is then broken down into three core structural categories: cavity components, guide components, and fastening components, and assigned importance weights of 0.6, 0.3, and 0.1 respectively. The cumulative wear of each component is then summed according to these weights to obtain the weighted value of cumulative wear. This enables precise quantification of wear and tear on multiple components, avoiding the one-sidedness of single-dimensional evaluation.

[0029] Then the cumulative value of erosion in the storage environment was calculated. Cumulative value of maintenance and repair gains Cumulative values ​​of erosion in the storage environment. Based on time-series data of warehousing environment in a standardized dataset, the suitable environmental thresholds for mold storage are set as follows: temperature 15-25℃, relative humidity 40%-60%, and dust concentration ≤0.5mg / m³. 3 Hourly statistics were compiled on the duration and extent of environmental parameters exceeding the suitable range. The erosion amounts due to excessive temperature and humidity, dust, and corrosive gases were calculated separately. These values ​​were then weighted and summed using coefficients of 0.5, 0.3, and 0.2 to obtain the hourly environmental erosion value. This summation was then applied over the entire storage period after the molds were placed in the warehouse to obtain the cumulative storage environmental erosion value. The erosion coefficient in high-temperature and high-humidity scenarios is 3.2 times that in normal-temperature and normal-humidity scenarios, fully quantifying the rust and aging damage of molds caused by the storage environment and solving the problem of traditional solutions ignoring health degradation during the storage stage. (Regarding the cumulative value of maintenance and repair gains...) The system quantifies and grades the maintenance and repair results based on the maintenance type and repair effect in the maintenance records: daily cleaning and maintenance is scored as 2 points per instance, blade grinding and repair is scored as 8 points per instance, cavity welding repair is scored as 20 points per instance, and core component replacement is scored as 35 points per instance. The gain value of all maintenance records is accumulated one by one. At the same time, a gain limit constraint is set, that is, the health after maintenance and repair shall not exceed 95% of the initial health, to avoid overestimation of health due to over-repair.

[0030] Furthermore, after quantifying all factors, the values ​​are substituted into the dynamic decay assessment model for mold health to calculate the real-time health level and dynamically determine the health grade. The expression for the dynamic decay assessment model for mold health is: ; This allows us to calculate the current real-time health status of the mold. Simultaneously, four threshold ranges for health levels are preset: when real-time health is ≥90, the health level label is judged as excellent; when 80≤real-time health is <90, it is judged as good; when 60≤real-time health is <80, it is judged as a warning; when real-time health is <60, it is judged as scrap. In this embodiment, an event-triggered dynamic update mechanism is also established. When the mold completes a production batch and returns to the warehouse, completes a maintenance operation, or the storage environment experiences 24 consecutive hours of exceeding the standard, the health level is immediately recalculated and the health level label is updated synchronously, realizing real-time dynamic tracking of health status.

[0031] Finally, based on the real-time health results, the predicted remaining service life was calculated. Health degradation data from nearly 50 production batches of the mold were extracted from the standardized dataset, and the average health degradation rate corresponding to a unit of effective stamping cycles was calculated using the sliding time window method. Simultaneously, nearly 180 days of idle storage data were extracted, and the average health degradation rate corresponding to a unit of storage time was calculated. Combining the current real-time health value with the scrap threshold of 60, the remaining health margin was calculated. Then, by combining the monthly production plan stamping volume and the monthly average storage time of the mold, the attenuation ratio during the production and storage idle stages is calculated respectively. Finally, the remaining number of stampings and the remaining usable time of the mold are calculated, which together constitute the predicted value of the remaining service life.

[0032] It should be noted that step 200, by constructing a dynamic health degradation assessment model integrating four dimensions—stamping fatigue, wear and tear, environmental erosion, and maintenance and repair—breaks through the limitations of traditional mold management that relies solely on usage frequency and manual experience to judge health status. This achieves precise quantitative calculation and dynamic determination of mold health levels. By distinguishing the degradation patterns in production and warehousing scenarios, it completes multi-factor refined prediction of remaining service life, promoting the upgrade of mold maintenance mode from fixed-cycle maintenance to predictive maintenance. This not only avoids cost waste caused by over-maintenance but also prevents the risk of production defects and mold scrapping due to untimely maintenance, comprehensively improving the control accuracy and operational economy of the entire mold life cycle.

[0033] In the specific implementation process, step 300 collects and marks the three-dimensional coordinates (lane number, floor number, column number), effective capacity size, rated load-bearing weight, functional area (ordinary storage area, maintenance buffer zone, pending processing area, complete set storage area) of each storage location in the automated warehouse, as well as the path distance to the nearest entry / exit station, to construct a complete warehouse space topology dataset; at the same time, it collects the entry / exit records of all molds in the past 30 days, and aggregates them according to the mold RFID code to obtain the cumulative number of storage and retrieval times, storage and retrieval time distribution, and complete set retrieval frequency of each mold, etc., as trajectory statistics.

[0034] In this embodiment, after completing the basic data preparation, two core efficiency-related adaptation factors are calculated sequentially: the access frequency adaptation factor and the health level adaptation factor. Regarding the access frequency adaptation factor... First, the cumulative number of storage and retrieval times for each mold over the past 30 days is normalized using a maximum-minimum method to obtain a normalized frequency value ranging from [0,1]. Simultaneously, the path distance from each storage location to the inbound / outbound platform is reverse-normalized; the closer the distance, the higher the location convenience value, also mapped to the [0,1] interval. The storage and retrieval frequency adaptation factor is the product of the normalized frequency value and the location convenience value. The factor value is maximized when matching high-frequency molds with storage locations near the platform, thus ensuring the operational efficiency of high-frequency storage and retrieval molds. (Health level adaptation factor...) Based on the four-level health status output in step 200, the corresponding storage area preferences are matched as follows: Excellent-level molds are preferentially assigned to high-rise, distant ordinary storage areas, with a corresponding storage location adaptation benchmark value of 1.0 and a low-rise maintenance area benchmark value of 0.4; Good-level molds are universally adapted to all areas, with a uniform benchmark value of 0.8; Warning-level molds are forcibly matched to low-rise, near-maintenance workstation dedicated storage areas, with a corresponding area benchmark value of 1.0 and a normal high-rise area benchmark value of 0.3; Scrap-level molds can only be assigned to dedicated storage areas to be processed, and the adaptation value of other storage locations is directly set to 0, thus achieving accurate matching between health status and storage resources.

[0035] In this embodiment, refined calculations are performed simultaneously on the spatial carrying capacity adaptation factor and the complete set association adaptation factor. Regarding the spatial carrying capacity adaptation factor... First, a hard constraint check is performed on the mold and candidate storage location: if the mold's dimensions exceed the storage location's effective tolerance, or the individual unit's weight exceeds the storage location's rated load capacity, the adaptation factor for that combination is directly set to 0, excluding infeasible allocation schemes; if the hard constraints are met, the storage location space utilization rate and weight utilization rate are further calculated. When the space utilization rate is in the optimal range of 60%-85%, the adaptation factor is set to its maximum value of 1.0. When the utilization rate is too low or too high, it decreases linearly, maximizing space reuse efficiency while ensuring storage security. For the set-related adaptation factor... The system extracts the set identification field from the basic attributes of molds in the standardized dataset to confirm the set mold group ID to which the current mold belongs, and obtains the set of storage location coordinates of all components that have been allocated within the same group. For the current candidate storage location, the Manhattan distance rule of the automated storage system is used to calculate its path distance to each allocated storage location in the same group, which is the Manhattan distance within the horizontal layer plus the vertical layer difference conversion distance, with each vertical layer converted to an equivalent horizontal distance of 2 meters. Then, the arithmetic mean of all path distances is calculated, and the average distance of all candidate storage locations is normalized. The smaller the average distance, the higher the adaptation factor value. If the current mold is the first component allocated in the set, the adaptation factor is uniformly set to a baseline value of 0.5, and is recalculated when subsequent components are allocated. This ensures that all components of the same set mold are stored together, thereby significantly reducing the handling cost of set retrieval.

[0036] In this embodiment, after completing all single-factor quantization, the location fit between a single set of molds and storage locations is calculated through multi-factor weighted summation, and a global integer programming solution model is constructed. The formula for calculating the location fit is: ; in , , , The weight coefficients for the corresponding factors are determined using the analytic hierarchy process (AHP) combined with warehouse operation objectives. The default values ​​are 0.35, 0.25, 0.2, and 0.2, respectively. These weights can be dynamically adjusted based on different operational strategies such as efficiency priority or space priority. The objective function is to maximize the sum of the fit between all mold and storage location combinations, introducing decision variables of 0-1. This indicates whether mold i has been assigned to storage location k, with 1 indicating it has been assigned and 0 indicating it has not. Simultaneously, five types of constraints are superimposed to construct an integer programming model: 1) each mold is assigned exactly one storage location; 2) each storage location can store at most one mold; 3) hard constraints on spatial carrying capacity; 4) the vertical layer difference between storage locations of members in the same group does not exceed two layers; and 5) health level corresponds to storage area constraints. This ensures that the solution simultaneously meets the requirements of efficiency, safety, and business rules.

[0037] Furthermore, a branch-and-bound method is employed to globally solve the integer programming model, generating the optimal storage location allocation scheme. The solution process first relaxes the 0-1 integer programming problem into a linear programming problem, and uses the simplex method to obtain the theoretical upper bound of the objective function. Then, a decision variable with a non-integer solution is selected and decomposed into two subproblems: one with the variable set to 0 and the other to 1. The upper bound and feasible lower bound of each subproblem are then calculated. During the iterative traversal, if the upper bound of a subproblem is less than the currently known global optimal lower bound, the branch is pruned directly without further downward traversal, thus significantly reducing the solution scale. By recursively executing the branch, bound, and pruning loop, all feasible branches are eventually traversed to obtain the optimal allocation combination that maximizes the global fitness.

[0038] This embodiment also establishes a dynamic recalculation mechanism triggered by multiple events to achieve real-time adaptive updates of the storage location allocation scheme. Specifically, three types of trigger events drive the scheme iteration: The first type is triggered by mold status change. When the health level label changes abruptly, the corresponding mold and its group members are immediately marked as reassignment objects, triggering local recalculation. The layout adjustment can be completed without full solution, thus improving response speed. The second type is triggered by production plan adjustment. When a change in the production schedule for the next 7 days is received from the MES system, the predicted access frequency of each mold is recalculated, triggering full storage location reassignment to adapt to changes in production rhythm in advance. The third type is triggered by storage facility failure. When the storage location sensor detects abnormal shelves, continuously exceeding environmental parameters, or other failures, the faulty storage location is immediately blocked, and the original stored molds are included in the reassignment queue and prioritized for scheduling to backup storage locations.

[0039] It should be noted that step 300, by constructing a multi-factor weighted adaptation algorithm that integrates four dimensions—access efficiency, health adaptation, space utilization, and complete set collaboration—breaks through the limitations of the traditional fixed classification and static allocation model of mold warehousing. It achieves precise location adaptation for industry characteristics such as heavy-duty and irregularly shaped molds, complete set usage, and varying health states. Simultaneously, by combining an integer programming model with the branch and bound method, it ensures the global optimality of the location allocation scheme, avoiding the local optima defects of traditional greedy algorithms. Furthermore, it incorporates a dynamic reallocation mechanism triggered by multiple scenarios, which can respond in real time to changes in mold status, production plans, and the warehousing environment, continuously iterating and optimizing the warehousing layout. This significantly improves warehousing space utilization, shortens the access and handling path for complete sets of molds, and also meets the professional requirements for healthy mold storage.

[0040] In the specific implementation process, step 400 extracts the storage and retrieval requirements of all related molds for the next 24 hours from the production scheduling data, covering four task types: production outbound, completed return to warehouse, maintenance relocation, and scrap outbound. Each task is assigned a unique task number and simultaneously marked with attributes such as the corresponding mold RFID code, target warehouse location, task type, required online time, associated production batch quality level, and whether it belongs to a complete mold set. These are then uniformly included in the pending task pool. Simultaneously, the warehousing operation environment is digitally calibrated: based on the warehousing space topology data, an AGV driving road network topology map is constructed, marking all feasible paths, aisle widths, meeting points, and speed-limited areas; the rated operating speed, retrieval and placement cycle time, and service aisle range of each automated warehouse stacker crane are calibrated; and the current location, operating status, and remaining power of all AGVs and stacker cranes are collected in real time to form a dynamic equipment resource pool.

[0041] In this embodiment, after completing the basic data preparation, the priority value of a single access task is calculated using a multi-factor weighted formula. The priority calculation formula is as follows: ; in , , These are the weighting coefficients for production task urgency, quality requirement suitability, and path efficiency, respectively. The default values ​​are 0.5, 0.3, and 0.2, and they can be dynamically adjusted according to production operation strategies. Production Task Urgency Factor Based on the urgency metric of task delivery, the following method is used: The remaining buffer time is calculated with the task's required online time as the endpoint and the current time as the starting point. A reverse normalization method is used to calculate the factor value; the shorter the remaining buffer time, the higher the urgency. The value range is [0,1]. For tasks marked as extremely urgent, an additional urgency bonus of 0.3 is added, and the bonus value does not exceed the upper limit of 1.0. Quality requirement adaptation factor. For scheduling constraints specific to the mold industry, the matching degree between the quality precision requirements of production batches and the mold health level is quantified as follows: When the production task has high precision quality requirements and the corresponding mold health level is excellent, the matching factor takes the maximum value of 1.0; when the corresponding mold is at a good level, it takes 0.7; when it is at a warning level, it takes 0.2, and molds at the scrap level are forcibly prohibited from participating in production and outbound tasks; the matching factor for tasks with ordinary quality requirements is generally reduced by 0.2 baseline gradients. This mechanism ensures that high-quality production tasks prioritize the use of high-health molds, thus mitigating quality risks from the warehouse scheduling perspective. Path efficiency factor Based on the inverse normalization of the estimated path length at the beginning of the task, the following is obtained: calculate the estimated total path length from the current idle device location to the target storage location and then to the task endpoint. The shorter the length, the higher the factor value, with a value range of [0,1].

[0042] In this embodiment, all tasks are sorted in descending order of priority value to obtain an initial task sequence. Then, sequence constraint verification is performed based on the specific business rules for mold warehousing. First, a complete set task binding verification is performed: all outbound tasks belonging to the same complete mold set are identified through the complete set identifier field. All component storage and retrieval tasks within the same set are marked as bound execution units, forcibly maintained in a continuous sequence, and the execution order follows the process retrieval logic of "small parts first, then large parts; inserts first, then mold frames," prohibiting production waits caused by splitting complete sets of tasks. Second, a warehouse location and equipment availability verification is performed: the real-time status of the target warehouse location corresponding to each task is checked, excluding faulty warehouse locations and invalid tasks corresponding to occupied warehouse locations; the task load of a single piece of equipment is verified to ensure that the continuous operation interval of a single AGV and stacker crane meets equipment heat dissipation and safety specifications.

[0043] In this embodiment, based on a compliant initial task sequence, a genetic algorithm is used to globally optimize the transport paths of all tasks. The core objectives are to merge tasks in the same direction, avoid congested sections, and minimize the total transport path length. Specifically, firstly, sequence encoding is performed: the complete task sequence is mapped to a chromosome, with each gene position corresponding to the execution order of an access task. Bound sets of tasks are encoded as a single composite gene and cannot be separated. Secondly, a fitness function is constructed: the reciprocal of the total transport path length of all tasks is used as the core fitness index; the shorter the total path, the higher the fitness. Then, iterative evolution is performed: a roulette wheel selection method is used to select individuals with high fitness rankings to enter the next generation of the population; task sequence fragments of different individuals are exchanged using a single-point crossover operator, preserving the binding integrity of complete sets of tasks during the crossover process; and the positions of two unbound tasks are swapped using a random site mutation operator to expand population diversity. The iteration termination condition is set to 120 generations of cumulative iterations, or 25 consecutive generations where the optimal fitness of the population does not improve. Finally, the task sequence corresponding to the individual with the highest fitness in the population is selected as the path-optimized task sequence.

[0044] Furthermore, upon receiving an emergency production order insertion instruction from the MES system, the system first adds the newly added mold storage and retrieval task to the task pool, while simultaneously ensuring that currently executing tasks are not interrupted. Only all unstarted tasks and the newly added emergency task are rescheduled. The priority values ​​of all tasks are then recalculated according to the priority calculation formula to determine the sequence order of the new task. Subsequently, a local genetic optimization strategy is employed, replanning the paths of only the 15 tasks before and after the insertion position, rather than iterating through all tasks, thus significantly reducing computation time. If the new emergency task has the highest priority, it is directly inserted at the beginning of the execution sequence and prioritized for execution on the nearest idle equipment. If the insertion causes some existing ordinary tasks to exceed the delivery buffer threshold, an overdue warning is automatically generated and pushed to the production scheduler for manual confirmation of whether to adjust the schedule. In addition, when unexpected events such as equipment failure or warehouse location anomalies occur, local rescheduling is also triggered to quickly adjust the subsequent task sequence to ensure the stability of the overall operation rhythm.

[0045] After completing global optimization and dynamic adaptation, the final optimal access execution sequence is generated. Each task in the sequence contains complete execution information such as execution order number, assigned device number, estimated start time, estimated completion time, target storage location coordinates, action type, and path planning details.

[0046] It should be noted that step 400, by constructing a multi-constraint priority scheduling mechanism that integrates production urgency, quality adaptation, and path efficiency, breaks through the limitations of traditional warehouse scheduling that only sorts by a single dimension of time or path. In particular, the health and quality matching scheduling and the set task binding execution rules designed specifically for the characteristics of the mold industry achieve deep collaboration between warehouse scheduling and production quality and process requirements. Combined with global path optimization of genetic algorithms, it merges operations in the same direction and avoids aisle congestion, significantly improving the operating efficiency of warehouse equipment and reducing ineffective handling energy consumption. With the event-driven emergency order insertion and abnormal rescheduling mechanism, it can quickly respond to changes in production schedules and on-site emergencies, ensuring production flexibility and delivery stability, and improving the overall operating efficiency of mold storage and retrieval.

[0047] In the specific implementation process, step 500 first breaks down the complete storage and retrieval execution sequence into an AGV handling sub-sequence and an automated warehouse stacker crane sub-sequence based on the main equipment. It further breaks down a single storage and retrieval task into five standard action nodes: "equipment unloaded positioning, fork extension for picking up goods, load transfer and movement, target location placement, and positioning confirmation and reset." Each node is configured with preset target coordinates, standard execution time, rated operating speed, and standard posture parameters. Instructions are sent in real-time to the motion controller of the corresponding equipment via industrial Ethernet. After receiving the instructions, the equipment sends back a verification receipt to ensure that there are no packet losses or errors. For continuous storage and retrieval tasks of complete sets of molds, a timing buffer window for the docking of the AGV and the stacker crane is set. The start and stop cycles of both are dynamically adjusted based on real-time position prediction to avoid equipment waiting while unloaded.

[0048] Next, UWB positioning base stations were deployed throughout the warehouse, and positioning tags were installed on the AGV body, stacker crane platform, and mold body to output the three-dimensional spatial coordinates, instantaneous speed, and direction of motion data of each entity in real time. High-precision weight sensors were embedded in the four force points of the stacker crane forks and the AGV platform to collect the load weight and off-center torque data in real time, which were used to determine whether the mold was successfully picked up or placed and whether there was any center of gravity shift. Industrial vision recognition units were deployed at the warehouse entrance and AGV docking station in each aisle, in conjunction with edge computing terminals, to identify the three-axis attitude angles of the mold and the offset relative to the warehouse location in real time, and to verify the RFID code of the mold a second time to prevent the wrong mold from being picked up or placed.

[0049] Furthermore, based on the synchronously collected execution status data, multi-dimensional deviation quantification calculations are performed to generate a comprehensive execution deviation. Specifically, three types of core deviations are calculated: position deviation is calculated based on the target coordinates, using the Euclidean distance between the real-time UWB position and the target coordinates. For stacker cranes, horizontal travel deviation, vertical lifting deviation, and fork extension deviation are further distinguished and weighted with weights of 0.4, 0.3, and 0.3 respectively to obtain the total position deviation, which is then mapped to the [0,1] interval through maximum and minimum normalization; duration deviation is calculated based on the preset standard duration of each action node, using the percentage difference between the actual execution duration and the standard duration, which is normalized to reflect the degree of deviation in execution efficiency; attitude deviation is calculated based on the standard storage posture of the mold, using the weighted sum of the yaw angle, pitch angle, and roll angle deviations obtained from visual recognition, with a focus on increasing the weight of the yaw angle deviation, which is then normalized to obtain the attitude deviation value. Finally, the weighted sums are calculated using coefficients of 0.4 for position deviation, 0.35 for attitude deviation, and 0.25 for duration deviation to obtain the overall execution deviation.

[0050] Furthermore, based on the comprehensive execution deviation, a two-level threshold and three-level response mechanism is used to generate differentiated dynamic correction instructions. Specifically, a first threshold for mild deviation and a second threshold for severe deviation are preset, with normalized values ​​of 0.1 and 0.3 respectively in this embodiment. When the comprehensive execution deviation is less than the first threshold, it is determined to be a minor fluctuation within the normal operating range of the equipment. The system only stores the deviation data and corresponding working condition labels in the historical deviation database, without interfering with the normal operation of the equipment. The data is retained for subsequent model iteration and equipment accuracy calibration. When the comprehensive execution deviation is between the first and second thresholds, it is determined to be a moderate deviation that can be corrected online. The system immediately generates real-time correction instructions for the equipment: for position deviation, coordinate fine-tuning is sent to the motion controller to correct the travel path and stopping point through differential control; for posture deviation, the extension speed and lifting angle of the forks are dynamically adjusted until the mold posture returns to the standard range; for duration deviation, the equipment operating speed is dynamically adjusted within the safe rated range to compensate for time lag; the correction process continuously collects status data and performs closed-loop verification until the deviation falls back to within the first threshold. When the overall execution deviation exceeds the second threshold, it is determined to be a serious deviation with a safety risk. The system immediately issues an emergency stop command to the equipment, triggers mechanical braking and cuts off power output. At the same time, it pushes a manual intervention warning to the on-site operation and maintenance terminal, along with the deviation value, real-time location and on-site visual screenshot.

[0051] It should be noted that step 500 constructs a complete closed-loop control system for the warehousing execution process through multi-source heterogeneous execution status collection, multi-dimensional weighted comprehensive deviation assessment, and a dynamic correction mechanism with graded responses. This breaks through the limitations of traditional mold warehousing, which is characterized by "heavy scheduling and planning, light process control, and no real-time feedback." Targeting the industry characteristics of heavy-duty, high-value, and high-precision mold storage and retrieval, it achieves a full-gradient response from position fine-tuning and attitude correction to abnormal shutdown. This not only improves the positioning accuracy and operational safety of storage and retrieval operations and reduces the risk of mold collision damage and equipment failure, but also reduces unnecessary manual intervention and equipment downtime through automatic real-time correction of moderate deviations, thereby improving overall operational efficiency.

[0052] like Figure 2 As shown in the figure, this embodiment of the invention also provides a mold storage management system based on big data analysis, including: The data preprocessing module is used to collect multi-source heterogeneous data throughout the entire life cycle of the mold through a multi-source acquisition terminal, and to perform standardized cleaning and association mapping to obtain a standardized dataset. The multi-factor calculation module is used to perform multi-factor decay calculations on the standardized dataset through the dynamic decay assessment model of mold health, and obtain the health level label and remaining service life prediction value of each mold. The allocation scheme calculation module is used to calculate the storage location fit degree based on health level labels, remaining useful life prediction values, storage space topology data and historical access trajectory data through a multi-factor weighted adaptation algorithm, so as to obtain the optimal storage location allocation scheme. The sequence optimization module is used to sort and plan the tasks based on the production scheduling task data, the optimal warehouse location allocation scheme and the health level label through a multi-constraint priority scheduling mechanism, so as to obtain the optimal storage and retrieval execution sequence. The feedback correction module is used to perform storage and retrieval operations through the warehouse execution equipment based on the optimal storage and retrieval execution sequence, while simultaneously collecting execution status data in real time and calculating deviations to obtain dynamic correction instructions.

[0053] The beneficial effects of this invention are as follows: 1) A five-level hierarchical data mapping chain based on RFID coding was constructed, which broke down the data barriers between warehousing, production MES, maintenance management, and quality inspection systems from the bottom layer. It realized the full-link structured association and two-way traceability from the mold body, components, usage process, maintenance records to quality association, and solved the problems of data fragmentation, discrepancy between accounts and actual situation, and poor traceability in traditional solutions. 2) By constructing a dynamic decay assessment model for mold health that integrates four dimensions—stamping fatigue, wear loss, environmental erosion, and maintenance and repair—and combining refined calculation logic such as multi-component wear weighting, environmental erosion classification and quantification, and maintenance gain upper limit constraints, it not only avoids redundant costs caused by excessive maintenance, but also prevents the risk of production defects and premature mold scrapping caused by untimely maintenance. This comprehensively improves the control accuracy and operational economy of the mold throughout its entire life cycle and reduces the total life cycle operating cost. 3) A multi-factor weighted adaptation algorithm that integrates access frequency, health level, space capacity, and complete set association is adopted. An integer programming solution model is constructed with the goal of maximizing the total global warehouse location adaptation. The global optimal warehouse location allocation scheme is obtained by solving the branch and bound method. At the same time, combined with a dynamic reallocation mechanism triggered by three types of events such as health status change, production plan adjustment, and warehouse location failure, it can respond to business changes in real time and continuously iterate and optimize the warehouse layout, improve warehouse space utilization, significantly shorten the storage and retrieval path of complete sets of molds, and significantly improve warehouse space utilization and storage efficiency. 4) Construct a multi-constraint priority scheduling mechanism that integrates the urgency of production tasks, the adaptability of quality requirements, and the efficiency of paths. Combined with genetic algorithms, it realizes global material handling path optimization, which not only significantly improves the operating efficiency of warehousing equipment and reduces the energy consumption of ineffective handling, but also realizes deep collaboration between warehousing scheduling and production scheduling and quality requirements, ensuring the production delivery rhythm and product quality stability. 5) By collecting execution status data in real time through UWB positioning, weight sensing, and visual recognition multi-source units, and calculating the comprehensive execution deviation through multi-dimensional weighted calculation of position, duration, and posture, a differentiated correction mechanism with two-level thresholds and three-level responses is adopted. This achieves full-gradient control from recording and retaining minor deviations, automatic correction at the equipment end for moderate deviations to early warning of safe shutdown for severe deviations. This breaks through the limitations of traditional warehousing, which is "heavy on planning, light on control, and no feedback", improves the stability of system operation and the speed of abnormal response, and reduces safety risks such as mold collision damage and equipment failure.

[0054] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0055] Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. Furthermore, those skilled in the art will recognize that, based on the ideas of this invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A mold storage management method based on big data analysis, characterized in that, Includes the following steps: Multi-source heterogeneous data is collected from multiple sources throughout the entire life cycle of the mold using a multi-source acquisition terminal, and then standardized cleaning and association mapping are performed to obtain a standardized dataset. The standardized dataset was subjected to multi-factor decay calculations using a dynamic decay assessment model for mold health, resulting in a health level label and predicted remaining service life for each mold. The optimal storage location allocation scheme is obtained by calculating the storage location fit degree of the health level label, the remaining useful life prediction value, the storage space topology data and historical access trajectory data through a multi-factor weighted adaptation algorithm. The optimal storage and retrieval execution sequence is obtained by sorting and planning the production scheduling task data, the optimal warehouse location allocation scheme and the health level label through a multi-constraint priority scheduling mechanism. Based on the optimal access execution sequence, access operations are performed through the warehouse execution equipment, and at the same time, execution status data is collected in real time and deviation is calculated to obtain dynamic correction instructions.

2. The mold storage management method based on big data analysis according to claim 1, characterized in that, The multi-source data acquisition terminal includes: RFID reader / writer, environmental sensor, production MES system, maintenance management system, and quality inspection system; the multi-source heterogeneous data includes: mold basic attribute data, warehousing environment data, production usage data, maintenance record data, and product quality data.

3. The mold storage management method based on big data analysis according to claim 1, characterized in that, Multi-source heterogeneous data is collected throughout the entire lifecycle of the mold using a multi-source acquisition terminal. This data is then standardized, cleaned, and correlated to obtain a standardized dataset, including: The missing points of continuous time series data in the multi-source heterogeneous data are fitted and filled by Lagrange interpolation, and the missing items of discrete attribute data in the multi-source heterogeneous data are filled with default values ​​by rule matching to obtain the completed dataset. The cleaned dataset is obtained by removing outliers from the completed dataset using the 3σ criterion. The normalized dataset is obtained by linearly transforming all numerical data of different dimensions in the cleaned dataset using the max-min normalization method. Using the unique RFID code of each mold as the primary key identifier, the associated identifier fields of the mold basic attributes, component list, production and usage records, maintenance and repair records and product quality data in the normalized dataset are extracted respectively. Based on the associated identifier field, a hierarchical mapping relationship is constructed to obtain a data mapping chain; the data mapping chain has a five-level hierarchical structure, with the following hierarchical order: mold body layer, component layer, usage record layer, maintenance record layer, and quality association layer; The data mapping chain is bound and stored with the normalized dataset to obtain the standardized dataset.

4. The mold storage management method based on big data analysis according to claim 1, characterized in that, The expression for the dynamic decay assessment model of mold health is: ;in, This represents the current real-time health status of the mold. This is the initial health baseline value after the mold has passed factory acceptance. The cumulative effective number of stampings for the mold. The weighted value of the cumulative wear of the mold. This represents the cumulative value of erosion in the storage environment. This represents the cumulative gain from maintenance and repair. , , and All of these are weighting coefficients.

5. The mold storage management method based on big data analysis according to claim 4, characterized in that, Multi-factor decay calculations are performed on the standardized dataset to obtain the health level label and remaining service life prediction value for each mold, including: The real-time health value is matched with a preset threshold range to obtain the health level label; the health level label includes: excellent, good, warning, and obsolescence. Based on the historical production data in the standardized dataset, the average health decay rate of the mold per unit time is calculated. Combined with the real-time health value and the scrap threshold, the remaining number of stamping cycles and the remaining usable time of the mold are estimated to obtain the predicted value of the remaining service life.

6. The mold storage management method based on big data analysis according to claim 1, characterized in that, A multi-factor weighted adaptation algorithm is used to calculate the storage location fit based on the health level label, the predicted remaining useful life, the storage space topology data, and the historical access trajectory data, to obtain the optimal storage location allocation scheme, including: The location fit is calculated using a multi-factor weighted fit algorithm; the formula for calculating the location fit is as follows: ;in, For access frequency adaptation factor, As a health level adaptation factor, As a spatial carrying capacity adaptation factor, For a complete set of related adaptation factors, , , and All are weighting coefficients; The objective function is to maximize the sum of the storage location fit of all mold and storage location combinations, and a preset constraint condition is added to the objective function to construct an integer programming solution model. The branch and bound method is used to solve the integer programming model to obtain the optimal allocation combination. When the health level label changes, the production plan is adjusted, or the storage location fails, a real-time recalculation is immediately triggered to update the allocation scheme and obtain the optimal storage location allocation scheme.

7. The mold storage management method based on big data analysis according to claim 6, characterized in that, The calculation process for the complete set of related adaptation factors includes: The complete set identification field identifies the complete set of molds to which the current mold belongs, and obtains the allocated storage location information of all mold components in the same set; Calculate the average path distance between the current candidate storage location and all allocated component storage locations in the same group based on the allocated storage location information, and determine the complete set association adaptation factor based on the average path distance.

8. The mold storage management method based on big data analysis according to claim 1, characterized in that, A multi-constraint priority scheduling mechanism is used to sort and plan the production scheduling task data, the optimal storage location allocation scheme, and the health level labels to obtain the optimal access execution sequence, including: The priority value of a single access task is calculated using the following formula: ;in, As a factor of production task urgency, For quality requirement adaptation factors, This is the path efficiency factor. , and All are weighting coefficients; All pending access tasks are initially sorted from high to low according to the priority values ​​to obtain an initial task sequence. The genetic algorithm is used to globally optimize the transport path corresponding to the initial task sequence, merge tasks in the same direction, avoid congested road segments, and minimize the total transport path length to obtain an optimized task sequence. When an emergency production order insertion instruction is received, the priority values ​​of all pending tasks and the newly added emergency task are recalculated, the sorting results are updated, and the subsequent task sequence is adjusted to obtain the optimal access and execution sequence.

9. The mold storage management method based on big data analysis according to claim 1, characterized in that, Based on the optimal access execution sequence, access operations are performed through the warehouse execution equipment, while execution status data is collected in real time and deviations are calculated to obtain dynamic correction instructions, including: The optimal access execution sequence is broken down into single-step equipment action instructions and sent to the AGV handling equipment and the automated warehouse stacker to perform access operations; The execution state data is obtained by collecting position data, weight data, and attitude data during the execution process through a UWB positioning unit, a weight sensor, and a visual recognition unit, respectively. The position deviation, duration deviation, and attitude deviation are calculated separately, and the overall execution deviation is obtained by weighted summation. When the overall execution deviation is less than a preset first threshold, only the deviation data is recorded and no correction instruction is generated; when the overall execution deviation is between the first threshold and a preset second threshold, a real-time correction instruction is generated on the device side to adjust the handling speed and pick-up / placement position; when the overall execution deviation is greater than the second threshold, the device operation is immediately suspended and a manual intervention warning is triggered.

10. A mold storage management system based on big data analysis, characterized in that, include: The data preprocessing module is used to collect multi-source heterogeneous data throughout the entire life cycle of the mold through a multi-source acquisition terminal, and to perform standardized cleaning and association mapping to obtain a standardized dataset. The multi-factor calculation module is used to perform multi-factor decay calculation on the standardized dataset through the mold health dynamic decay assessment model to obtain the health level label and remaining service life prediction value of each mold. The allocation scheme calculation module is used to calculate the storage location adaptation degree of the health level label, the remaining useful life prediction value, the storage space topology data and historical access trajectory data through a multi-factor weighted adaptation algorithm, so as to obtain the optimal storage location allocation scheme. The sequence optimization module is used to sort and plan the production scheduling task data, the optimal storage location allocation scheme and the health level label through a multi-constraint priority scheduling mechanism to obtain the optimal storage and retrieval execution sequence. The feedback correction module is used to perform storage and retrieval operations through the storage execution equipment based on the optimal storage and retrieval execution sequence, and at the same time collect and calculate the deviation of the execution status data in real time to obtain dynamic correction instructions.

Images