Work order material intelligent tracking and dynamic updating system based on multi-source data fusion

Abstract

The invention relates to a work order material intelligent tracking and dynamic updating system based on multi-source data fusion. The system comprises an acquisition module, a tracking module, an updating module, a man-machine interaction module and a storage module, the acquisition module is used for acquiring work order information from a bill of material, scheduling information and inventory and supplier data, and constructing a material data model; the tracking module monitors material inventory, consumption and material shortage risks based on the model, and realizes automatic state updating and graphical path recording in combination with bar code recognition and ambiguity analysis; the updating module predicts material requirements through a machine learning algorithm, optimizes a distribution plan and outputs a dynamic report; the interaction module supports multi-format data import, unstructured field classification and interrupted operation control; the storage module ensures credible synchronization and tamper-proofing performance of data among different systems through a block chain technology; according to the invention, high-efficiency management and dynamic allocation of work order materials can be realized in an environment without an MES system, and the production transparency and the resource utilization efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of material management, and particularly to an intelligent tracking and dynamic updating system for work order materials based on multi-source data fusion. Background Art

[0002] In the prior art, work order material management usually relies on enterprise resource planning (ERP) or manufacturing execution system (MES) for work order generation, material allocation, and inventory management. These systems achieve unified management of information such as bill of materials (BOM), inventory status, and production scheduling through an integrated database, and support barcode scanning, label printing, and partial data tracking functions. In the case of some small and medium-sized manufacturing enterprises lacking the above systems, work order management and material counting are also carried out through manual forms or simple software, supplemented by barcode devices for partial auxiliary operations.

[0003] However, there are multiple problems in the prior art: firstly, there is a lack of a fusion mechanism between multi-source data, and work order data is often scattered in multiple systems or tables, making it difficult to construct a unified data view; secondly, there are problems of ambiguous matching and inconsistent material numbers in barcode recognition, resulting in status update errors; thirdly, there is a lack of an effective demand forecasting and allocation optimization mechanism, which is likely to cause material shortages or excessive inventory; fourthly, some systems cannot support the import of unstructured data and interrupted continuation, with insufficient operation flexibility; fifthly, cross-system data synchronization lacks reliable guarantee, and there are risks of data errors, omissions, and tampering.

[0004] Therefore, there is an urgent need for a new work order material management solution that can fuse multi-source data and achieve intelligent tracking and dynamic updating of material status. Summary of the Invention

[0005] The present application provides an intelligent tracking and dynamic updating system for work order materials based on multi-source data fusion to improve production transparency and resource utilization efficiency.

[0006] The present application provides an intelligent tracking and dynamic updating system for work order materials based on multi-source data fusion, including: A collection module for collecting work order data from a bill of materials, production scheduling, inventory records, and supplier databases, and constructing a unified material data model through a standardized protocol and semantic algorithm; A tracking module for real-time tracking of material status based on the material data model, including inventory, consumption, and material shortage risks, and matching barcodes with material numbers through an integrated barcode recognition and ambiguity resolution unit to achieve automatic status update and graphical path recording; An update module for receiving material status data, predicting material requirements based on machine learning algorithms in combination with historical records and usage frequencies, and adaptively adjusting material allocation to generate a dynamic report associated with the original work order; An interaction module for data interaction with the acquisition module, tracking module, and update module, displaying the current material data model, inventory counting records, and dynamic reports; supporting the simultaneous import of the original bill of materials and historical inventory counting tables, completing the classification and mapping of unstructured data based on a field recognition mechanism, and having the functions of resuming interrupted inventory counting and printing dynamic reports and outputting labels based on user permissions; A storage module for storing the material data model, inventory counting records, and dynamic reports in a distributed database, and ensuring data consistency and immutability across systems through a blockchain synchronization mechanism.

[0007] The beneficial effects of this application mainly include: (1) It can construct a unified material data model, realize the fusion and standardized processing of multi-source heterogeneous data from bills of materials, scheduling plans, inventory records, supplier databases, etc., and significantly improve the integrity and consistency of work order data. (2) By integrating barcode recognition and ambiguity resolution mechanisms, it can automatically match material numbers and update material status in real time, significantly improving the accuracy of inventory counting and avoiding identification errors caused by multi-code conflicts in traditional barcode systems. (3) Introducing machine learning algorithms to predict material requirements and adaptively adjust material allocation, dynamically generating reports based on historical inventory counting and usage frequencies, effectively reducing the risk of material shortages and overstocking problems, and enhancing the responsiveness of production plans and resource utilization rates. (4) It supports resuming interrupted operations, importing unstructured data, and outputting with permission control, and combines the blockchain synchronization mechanism to ensure the security, credibility, and immutability of data in a distributed environment, enhancing the flexibility of the system and data control capabilities. Brief Description of the Drawings

[0008] Figure 1 is a schematic diagram of a work order material intelligent tracking and dynamic update system based on multi-source data fusion provided by the first embodiment of this application. Detailed Embodiments

[0009] Many specific details are set forth in the following description in order to provide a thorough understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the connotation of this application. Therefore, this application is not limited by the specific embodiments disclosed below.

[0010] The first embodiment of this application provides a work order material intelligent tracking and dynamic update system based on multi-source data fusion. Please refer to Figure 1 , which is a schematic diagram of the first embodiment of this application. The following will be combined with Figure 1 to describe in detail a work order material intelligent tracking and dynamic update system based on multi-source data fusion provided by the first embodiment of this application.

[0011] The intelligent tracking and dynamic update system for work order materials based on multi-source data fusion includes a collection module 101, a tracking module 102, an update module 103, an interaction module 104, and a storage module 105.

[0012] The collection module 101 is used to collect work order data from a bill of materials, production schedule, inventory records, and supplier database, and construct a unified material data model through a standardized protocol and semantic algorithm.

[0013] The core function of the collection module 101 is to complete the multi-source collection and fusion modeling of work order-related data. Its technical implementation includes, but is not limited to, the following specific steps and mechanisms, ensuring that those skilled in the art can fully implement it accordingly: First, the collection module 101 connects to the bill of materials (BOM), production scheduling system, inventory management records, and supplier database respectively by setting a unified data access interface. When each data source is accessed through the interface, the system will call a preset data parsing template matching the data source. For example, for data file formats based on CSV, XLSX, JSON, or SQL, the corresponding parser components will be called to convert the original data into an intermediate format data object. This parsing process not only completes field recognition but also includes unified specification processing of date formats, quantity units, character encodings, etc., ensuring subsequent compatibility.

[0014] Second, after the conversion of the intermediate format data is completed, the collection module 101 further executes the data standardization protocol. Based on the field mapping table and unit conversion table built into the system, this protocol maps information with the same semantics but different field names and structures in different data sources to a standard field structure. For example, fields such as "Item Code", "Material ID", and "PartNumber" are all unified as "Material_ID", and "Planned Qty" and "Expected Amount" are unified as "Planned_Quantity", and automatic conversion of units such as "kg" and "g" is performed. In addition, for missing fields or irregular data, this module generates exception logs through a data verification mechanism and records the fields to be filled, prompting the user for manual confirmation or default filling.

[0015] After data normalization is completed, the acquisition module 101 calls the built-in semantic association algorithm to fuse multi-dimensional information from different data sources based on keyword fields such as material number, supplier code, timestamp, and category label. Specifically, the system adopts a strategy combining hash mapping and entity alignment to align the primary keys of data entries that are semantically equivalent but from different sources. If field conflicts are found in different sources, such as inconsistent planned quantities under the same material number, the data source with a higher user-set priority is preferentially retained, and the conflict status and change records are marked in the data model.

[0016] Taking the work order management of a manufacturing enterprise as an example, the system needs to integrate data from two types of data sources: one is the BOM table exported from the ERP system (source A), and the other is the material delivery schedule provided by the supplier (source B). In both types of data, there is a field labeled "material number", but in source A, this field is "Part Number", and in source B, it is "Material Code". The system automatically recognizes their semantic equivalence through the field mapping table and maps them to the standard field "Material_ID".

[0017] During the integration process, the system first generates a hash value for each record based on the "Material_ID" field to quickly match record entries from different sources. For example: Record from source A: Material_ID = 123456, Planned_Quantity = 500; Record from source B: Material_ID = 123456, Planned_Quantity = 520; The system detects that there are records with the same Material_ID = 123456 in the two sources, but there is a conflict in the "Planned_Quantity" field. At this time, the system checks the user's preset priority setting and finds that the priority of source A is 1, which is higher than the priority of source B, which is 2.

[0018] Therefore, the system writes the Planned_Quantity = 500 from source A as the final value into the unified material data model. At the same time, a conflict flag field (such as Conflict_Flag = True) is added to this record entry, and an attached change record sub-table records the following information: Conflicting field: Planned_Quantity; Original value from source B: 520; Source of the current value: Source A; Conflict handling timestamp and operation description: e.g., "On May 5, 2025, 14:22, retain the data from source A according to the priority setting." Through this specific processing flow, when the system integrates data that is semantically equivalent but has inconsistent values, it not only achieves primary key alignment and conflict decision-making but also has complete tracking and auditing capabilities, ensuring the traceability of subsequent data tracking and responsibility determination. This strategy can be stably adapted to various actual factory data fusion scenarios, helping to improve the intelligence level and system robustness of work order data management.

[0019] Finally, the acquisition module 101 constructs a unified material data model based on the standardized data entries. This model is a structured multi-table associated database structure, including at least the following fields: Material_ID, Material_Name, Planned_Quantity, Available_Inventory, Supplier_ID, Schedule_ID, Status_Flag, etc., and supports foreign key association to the production work order table and the supplier information table. The material data model is stored in the form of JSON Schema or relational table structure and can be directly read and updated by subsequent modules of the system (such as the tracking module 102 and the update module 103).

[0020] To adapt to dynamic business scenarios, the acquisition module 101 supports a periodic data refresh mechanism, which can periodically re-collect specified data sources according to the time strategy set by the user (such as every 30 minutes, daily synchronization) and compare and merge them with the current data model. For external systems such as ERP / SCM systems, incremental data synchronization can also be achieved through API interface calls, supporting WebSocket or RESTful standards.

[0021] Through the above clear steps and mechanisms, the acquisition module 101 not only realizes the real-time access, semantic fusion, and modeling normalization of multi-source data but also provides a highly consistent and high-quality underlying data foundation for subsequent material status tracking, dynamic prediction, and report generation. Its standardization and semantic analysis capabilities enable it to adapt to work order data in different formats from different suppliers and different industries, with high generality and scalability.

[0022] The tracking module 102 is used to track the material status in real time based on the material data model, including inventory, consumption, and shortage risk, and match barcodes with material numbers through the integrated barcode recognition and ambiguity resolution unit to achieve automatic status update and graphical path recording.

[0023] The function of the tracking module 102 is to perform a whole-process, real-time, and automated dynamic tracking of the status of various materials in the work order based on the material data model. Its implementation process involves multiple specific steps such as data reading, status calculation, barcode matching, and graphical output, and requires clear configuration and logical processing to ensure its complete feasibility.

[0024] When the system runs, the tracking module 102 first calls the unified material data model constructed by the acquisition module 101 to read the basic attributes of each material, including but not limited to material number (Material_ID), planned quantity, received quantity, issued quantity, supplier identifier, and work order number, etc. The core target status fields set by this module include current inventory, cumulative consumption, replenishment status, shortage risk level, etc., and its specific calculation logic is as follows: Current inventory = total received quantity - total issued quantity, Cumulative consumption = quantity already invested in the work order, and the shortage risk is determined by comparing with a threshold. For example, if the current inventory is less than 20% of the planned quantity, it is marked as a high-risk status.

[0025] To achieve automated status updates, the tracking module 102 integrates a barcode recognition and disambiguation resolution unit, which adopts a dual-channel design of image recognition and string recognition, and uses a high-precision scanning head or camera device to recognize the one-dimensional code or two-dimensional code pasted on the material packaging. The system first reads the barcode content and parses its corresponding fields (such as material number, batch number, etc.), and then matches the barcode with the material number field in the material data model. If the match is unique, the system automatically regards this scanning action as an "incoming" or "consumption" behavior, and automatically updates the increment of the inventory or consumption according to the material counting link where the user is located (such as pre-production material counting or post-production recycling).

[0026] If the read barcode corresponds to multiple possible material items (such as multiple material codes having a similar or nested relationship), the disambiguation resolution mechanism is triggered. The system searches for all matchable items in the material data model, and sorts and recommends them in combination with the current material status, batch records, and user operation context, prompting the user to select the most likely match item, or automatically selecting the item with the highest confidence level when permissions allow. After each match, the system writes the update operation into the material counting record form as a record item, and uses colors or marks in the interface to prompt that the status of this material has been updated.

[0027] In addition, the tracking module 102 generates a path-based record of all material status changes and visualizes them graphically, supporting querying of the material flow path through multi-dimensional filtering methods such as time axis, work order number, material name, etc. The path display structure is constructed based on the material life cycle status diagram, such as from "planned allocation" to "received" and then to "consumed" or "shortage". Each node is attached with a timestamp and operation source (such as scanning code, manual correction, system prediction, etc.), and has a historical status backtracking function.

[0028] The system also supports a batch status verification mechanism triggered by scheduled tasks, which can automatically scan the status of all materials under the current work order at specified time intervals, compare with the work order progress, identify material lag or overuse anomalies, and generate marks for subsequent analysis and processing by the update module 103.

[0029] In summary, through the deep integration of multi-source data drive, status logic calculation, barcode precise matching, and graphical path recording, the tracking module 102 realizes the dynamic monitoring and highly reliable tracking of the material status of work orders, forms a structured understanding of the entire process of material flow, provides real-time data support for intelligent allocation and shortage warning, and can be efficiently deployed and stably operated in the actual manufacturing process.

[0030] Furthermore, the tracking module is specifically used for: Collect the current barcode scanned by the user based on the barcode scanning event, and combine it with the previous barcode scanning records to construct the current barcode scanning sequence, and extract the combined features including the barcode arrangement order, time interval, and historical matching accuracy; According to the combined features of the barcode scanning sequence and the current material status of multiple material numbers that the barcode may correspond to in the material data model, identify the material status threshold field, and through the context association selection mechanism, determine the target material number most likely to match the barcode among multiple material numbers; Use the target material number as the input to drive the update operation of the material status, mark the matching decision path in the update record, and record the status evolution path of the material during the counting process as time series data; Based on the time series data, construct a multi-dimensional status graph including the status switching frequency, interruption position, and user intervention records, and display it in the form of a heat map on the system interface for the user to make a shortage risk prediction and decision reference.

[0031] During the actual operation of the system, first, the user uses a barcode scanning device to scan the barcode on the material package. After receiving the barcode scanning event, the tracking module takes the scanned content (i.e., the current barcode) as input data, and at the same time calls the previous barcode scanning records cached or stored to construct a current barcode scanning sequence. This sequence includes not only the list information of the current barcode and the previous barcodes, but also parameters such as the scanning timestamps and their intervals, and the confidence level of the material number matching during scanning. The system extracts feature vectors from this sequence, specifically including the arrangement order of barcodes, the time interval between consecutive scans, the historical matching success rate, and the number of consecutive matching failures. These features are used to determine the true ownership of the barcode in the subsequent process.

[0032] In an actual factory, different materials may have similar or overlapping barcodes (e.g., the encoding format of one-dimensional barcodes is repeated or two-dimensional barcodes contain nested information). Therefore, a single barcode may map to multiple candidate material numbers. At this time, the system identifies the "material status threshold field" based on the preset status information of each candidate material number in the material data model, such as the inventory status field, the field indicating whether the counting has been completed, the current number of references, etc. This field usually includes key indicators such as whether the inventory is zero, whether the last update is overdue, and whether it is in the manual confirmation state. Then, the system executes a context-related selection mechanism, that is, compares the extracted barcode combination features with these threshold fields, calculates the matching probability of each candidate material number with the current barcode, and selects the entry with the highest matching degree as the "target material number". This process can be achieved through a weighted scoring function, and the weights of each item in the scoring formula can be set by system training or operation and maintenance personnel and are allowed to be adjusted dynamically during operation.

[0033] When the target material number is determined, the system immediately uses this number as the entry point for the status update operation, calls the status update logic, and modifies the current status field of the material, such as updating the completed counting quantity, marking the scanned status, modifying the shortage risk level, etc. At the same time, the system records the decision-making path of this matching operation in the background update log, specifically including: the set of candidate material numbers, the matching probability score, the finally selected number, the referenced context information (such as the scanning sequence number and time interval), and the user ID, etc., to ensure that every change in the material status is traceable and the record of responsibility attribution.

[0034] Furthermore, the system records all the status evolution situations of the target material during the counting process as time series data. This time series includes: the timestamp of each status change, the status type (such as "scanned", "confirming", "shortage warning", "adjusted", etc.), the triggering method (system automatic recognition or manual intervention), and the previous and current status values, etc. This sequence can be stored as a timestamp-keyed structured array or database record, supporting fast retrieval and historical restoration.

[0035] Finally, the system constructs a multi-dimensional state graph based on the above time series data, where each node represents a specific state value, and the edges between nodes represent state transition paths. For each node, additional dimensional information is attached, such as the switching frequency (i.e., the number of times the state is repeated), whether there is user intervention (such as manual correction, forced submission), whether it is aborted due to an abnormal interruption, etc. This multi-dimensional state graph is used to generate a heat map display. The heat map uses different shades of color to represent different risk levels or abnormal densities and is integrated into the system interface to provide users with a visual risk identification assistance function. Users can dynamically switch the image view according to different filtering conditions (such as by material number, by work order number, by status label) to observe the dynamic trajectory and bottleneck risks of a certain material in the entire production process.

[0036] In summary, this tracking module not only realizes the barcode context semantic recognition that cannot be covered by traditional point material systems, but also models the material behavior as an interpretable, traceable, and interactive graphical process, significantly improving the system's intelligent recognition ability for barcode conflicts and state tracking problems in complex production environments.

[0037] Furthermore, the context correlation selection mechanism determines the target material number by calculating the matching scoring function between the barcode and the candidate material number to determine the target material number. The matching scoring function is defined as Formula 1 below: ;

[0038] where is the context consistency coefficient; is the material state difference degree; is the reference confidence; is the risk flag index; is the relative time activity; , , , and are weight coefficients, and their recommended values are 0.35, 0.25, 0.15, 0.15, and 0.1 respectively.

[0039] Among them, the context consistency coefficient is shown in Formula 2 below: ;

[0040] where represents the semantic vector angle between the candidate material number and the previous scanned material number ; represents the time interval between two scans; Represents the historical successful matching weight of the previous barcode scanning event, which is the ratio of the number of times the previous barcode has been successfully matched in the past to the total number of times the previous barcode has been scanned in the past; Is the time decay coefficient; Represents the number of barcodes recorded previously in the current barcode scanning sequence, that is, the number of consecutive valid barcode scanning events recorded by the system before the current user scans this barcode; The following combines specific examples to elaborate on each item in Formula 2: Context consistency coefficient Is used to measure the semantic consistency and operational timing correlation between "the candidate material number that the current scanned barcode may correspond to " and "the materials corresponding to the barcodes continuously scanned by the user before". Its essence is: if the candidate material number is semantically similar to the material number scanned by the user before, and the barcode scanning times are close, and the user rarely makes mistakes historically, then it is more likely to be the correct match for the current barcode.

[0041] : The number of barcodes continuously and successfully scanned by the user before the current barcode scanning. For example, if the user has just continuously scanned 3 materials (such as the motherboard, battery, and shell), then .

[0042] : The "semantic difference angle" between the candidate material number i and the jth previous material. It can be understood as whether two materials are similar in the bill of materials or classification model. For example, "charger" and "adapter" are semantically close and the angle is small; while "charger" and "screw" are completely different and the angle is large. In a simplified implementation, the "semantic difference angle" can be estimated by comparing the similarity levels of the candidate material number and the previous material number in dimensions such as the historical co-occurrence relationship in the material name, category label, or work order. Specifically, the system establishes a preset semantic level table to represent the semantic similarity level between common material pairs. If the names of two materials contain the same or approximate keywords (such as "battery", "adapter", "package", etc.), or they often appear continuously or share the same work station in the historical work order, a higher semantic consistency score is given, indicating a smaller difference; if there is no obvious association between them in terms of name, category, or usage process, the score is lower, indicating a larger semantic difference.

[0043] : The cosine value from the semantic angle, the larger the value, the more similar (the maximum is 1), and the smaller the value, the less relevant or even opposite (may approach 0).

[0044] : The time interval between the current barcode scan and the j-th barcode scan. The closer the two scan times are, the more likely they are to occur in the same assembly process and the stronger the correlation.

[0045] : The time decay factor, which controls the degree of influence of the time distance on the result. The longer the time interval, the lower the weight. The recommended value is 0.2.

[0046] : The historical matching confidence, that is, the recognition success rate of the system for the j-th barcode in the past. If the barcode is often misrecognized, the weight is low; if it is often recognized correctly, the weight is high.

[0047] Material status difference Is defined as the following formula 3: ;

[0048] Among them, Represents the candidate material number The value on the status field ; Is the value of the system's current reference material on the same field; Is the total number of comparison fields, Is a very small positive number to prevent division by zero, for example ; This indicator is used to measure the relative deviation degree of the candidate material from the current context state; The reference material refers to: In the process of barcode recognition or inventory counting, the system selects a known material number based on the context as the comparison reference material for the currently to-be-recognized barcode. It is usually the material that the user has just recognized or the material that the system considers to be most relevant to the current operation, and is used to assist in judging which candidate material the current barcode is more likely to correspond to.

[0049] Suppose the user is conducting inventory counting for a production work order, and the last scanned and confirmed material is "Power Adapter Model A". At this time, the user continues to scan the barcode, but the scanned barcode may correspond to multiple materials, such as "Power Adapter Model A", "Power Adapter Model B", or "Spare Battery".

[0050] Due to overlapping or incomplete barcode information, the system cannot directly confirm which material the newly scanned barcode corresponds to. Therefore, it uses the "Power Adapter Model A" that the user has just confirmed as the reference material, and compares the inventory status, whether it is locked, usage frequency and other fields of each candidate material to see which one is closer to the status of the reference material.

[0051] If the status characteristics of "Model A" and the current barcode are highly similar, while there are significant differences between "Model B" and the "spare battery" in terms of inventory or usage records, the system is more inclined to believe that the current scanned code still corresponds to "Model A".

[0052] Suppose the candidate material number A001 in the system is identified as one of the potential matching objects for the current scanned barcode, and the current context reference material is R123. The status fields selected for comparison are shown in Table 1 below: ; Table 1 Status Field Comparison Substitute each item into the formula: Field ;

[0053] Field ;

[0054] Field ;

[0055] Field 4: ;

[0056] Field ;

[0057] Calculate the average value (let ): ;

[0058] The calculated indicates that there is a 34% average relative deviation between the candidate material number A001 and the current reference material R123 in the key status fields, which can be used for horizontal comparison with other candidate materials to assist the context association selection mechanism in making an optimal matching judgment.

[0059] Reference confidence represents the frequency proportion of successful matching of the candidate material number in the recent scanning history, which is obtained by calculating the ratio of the number of hits to the total number of scans within a fixed time window.

[0060] Risk flag indicator is a binary parameter. If the current inventory of the candidate material is zero, the spot check record is abnormal, or it is in a locked state, the value is 1; otherwise, it is 0.

[0061] Time relative activity represents the ratio of the operation frequency of the candidate material within the current window to the average operation frequency of all candidate materials, and is used to measure its current activity level.

[0062] The system is based on the above scoring function Score all candidate material numbers, select the one with the highest score as the target material number for the current barcode, and record all parameter values and calculation paths as the update record of the tracking module for reference in status graph construction and manual review.

[0063] Furthermore, the tracking module is also used for: Based on the barcode scanning event, collect the current barcode scanned by the user, combine it with the previous barcode scanning records to construct the current barcode scanning sequence, extract the combined features including barcode arrangement order, time interval, and historical matching accuracy, and further determine whether there is an abnormal behavior pattern in the barcode scanning sequence where the time interval fluctuation exceeds the system preset threshold or the barcode arrangement order deviates from the production process logic model; When it is determined that the current barcode scanning sequence contains an abnormal behavior pattern, introduce the abnormal behavior mark as an input factor into the context - associated selection mechanism to adjust the weight distribution among multiple material numbers, and enhance the confidence influence of candidate material numbers with high historical matching accuracy and stable barcode scanning context in the determination of the target material number; Before driving the update operation of the material status with the target material number determined by the context - associated selection mechanism, add a trust control strategy for the updated status based on the abnormal behavior mark, specifically including marking this match as a to - be - verified status and restricting the update range of the material status until a subsequent supplementary barcode scanning event or user manual confirmation is received and then officially submitted to the material data model; While recording the state evolution path of the material during the material counting process as time - series data, record the path nodes intervened by abnormal behavior in the multi - dimensional status graph with significant markings, where the markings include path color change or risk flag superposition, and display it in the form of a heat map on the system interface for the user to make a shortage risk prediction and an error - correction decision based on barcode scanning anomalies.

[0064] In actual operation, the user completes the material counting action by scanning the material label (usually a one - dimensional code or a two - dimensional code). The tracking module will collect the current barcode scanned by the user in real - time and call the previous barcode scanning events recorded in the system to construct the current barcode scanning sequence. This barcode scanning sequence includes not only the content of the scanned barcode itself, but also the time stamp and the scanning order position of each barcode scanning. The system analyzes the time interval between barcode scanning events to determine whether the user is performing continuous material counting in a rhythm - compliant manner, and judges whether the current barcode scanning deviates from the reasonable order of the process flow by comparing it with the standard process material order.

[0065] For example, if the material numbers A1, A2, and A3 should be scanned sequentially in the standard process, but the current scanning order is A1, A3, A2, and there are significant fluctuations in the interval between the two scans (such as a sudden timeout after consecutive scans in a short period), the system will determine that the current scanning sequence has a time fluctuation exceeding the preset threshold or a deviation from the sequential logic, and then mark this scanning behavior as an abnormal behavior pattern.

[0066] This abnormal behavior mark is not used as an error handling termination operation, but is introduced as an input factor into the context - associated selection mechanism. This mechanism will adjust the matching weights between multiple material numbers by combining the combined features in the scanning sequence and the current material status in the material data model, and preferentially enhance the weight of candidate material numbers with high historical matching success rates, reasonable scanning sequences, and stable time intervals in the process of determining the target material number, thereby improving the overall confidence of the target matching.

[0067] After the system determines the target material number according to the context - associated selection mechanism, it does not immediately update the material status, but executes a credibility control strategy based on the abnormal behavior mark. This strategy includes marking the current status update as "to - be - verified status", that is, this update record is only temporarily stored in the logical layer and cannot be written into the formal material data model. The system will wait for the occurrence of a supplementary scanning event (such as subsequent scan confirmation) or manual confirmation of the matching result by the user. Only after meeting the credibility conditions will this status be formally submitted to ensure that the system does not mistakenly update the real material status data in case of anomalies.

[0068] Meanwhile, the tracking module will continue to record the state evolution path of this material during the counting process as time - series data over time, and mark the path nodes where abnormal behaviors intervene in the multi - dimensional state diagram. For easy visual recognition, the system uses significant visual identifiers for these abnormal path nodes in the interface heat map, such as using orange - red path lines, adding "abnormal" icons, or flashing borders, enabling users to clearly identify the potential risks generated by the automatic adjustment behaviors inside the system, and thus make more accurate interventions in the shortage risk assessment and on - site decision - making.

[0069] In summary, the tracking module not only realizes the dynamic perception of scanning behaviors and logical anomaly judgment, but also introduces an associated control and hierarchical verification mechanism at the levels of material number matching, status update, and graphical display, effectively improving the fault - tolerance ability, security, and interpretability of the system in the actual counting scenario.

[0070] The update module 103 is used to receive material status data, combine historical records and usage frequencies, predict material requirements based on machine - learning algorithms, and adaptively adjust material allocation to generate a dynamic report associated with the original work order.

[0071] The core function of the update module 103 is to implement the dynamic adjustment of the material prediction and distribution plan, and generate a dynamic report corresponding to the original work order data accordingly. The following will be a detailed description from four aspects: its input data source, prediction model construction, distribution strategy generation, and report output.

[0072] During the actual operation process, the update module 103 first continuously receives material status data from the tracking module 102. This data includes current inventory, consumed quantity, shortage risk level, and historical spot-check records, etc. At the same time, the update module can also call the original material data model and historical work order data provided by the acquisition module 101, including information such as past material usage quantity, delivery cycle, batch difference, and supply delay. All this information is uniformly loaded into the memory data structure of the update module to form a multi-dimensional input feature set, which is used to support subsequent prediction operations.

[0073] In terms of prediction model construction, the update module 103 presets a trainable machine learning algorithm, preferably based on regression trees, LSTM, or lightweight neural network architectures, to predict the demand change trend of each material in the next period of time according to the above features. The system can set a prediction window (such as the next 7 days, 30 days), and update the prediction results in real time according to the sliding window mechanism. Taking LSTM as an example, the system first segments the material usage curve in the historical work order by the time axis, then uses standardization and normalization methods to process the input features (such as unit usage, beat rate, delivery duration), and then inputs them into the LSTM neural network to train the model parameters, so as to obtain the predicted value of future demand.

[0074] After the prediction results are generated, the update module calculates whether there is a shortage for each material according to the difference between the current inventory and the predicted usage. If the shortage exceeds the set threshold (such as 10%), the allocation optimization process will be triggered. The system will call the adaptive allocation algorithm at this stage to adjust the "allocation quantity" field in the material data model in real time according to indicators such as the priority of each work order, supply cycle, and inventory turnover target. For example, when both Work Order A and Work Order B require the same material and the total demand exceeds the inventory, the system can allocate proportionally according to the priority or historical execution rate, and record the adjustment information into the dynamic version control module at the same time to ensure the traceability of the system's operations.

[0075] In the dynamic report output stage, the update module generates a comparative analysis report by comparing the aforementioned prediction results with the data structure after allocation and adjustment. The output content includes fields such as predicted usage, current inventory, recommended replenishment quantity, planned allocation ratio, shortage risk level, and recommended replenishment time. The system supports presenting the report in various forms such as tables and charts, and at the same time establishes an indexing relationship with the original work order number, enabling users to quickly retrieve the current material prediction status according to specific work orders. This report can be output as PDF, Excel, or directly pushed to the interaction module 104 for interface display.

[0076] The design of the entire update module features strong real-time performance, interpretability, and prediction adjustability, and is suitable for local chemical work order material allocation optimization in small and medium-sized manufacturing enterprises without the support of ERP / MES platforms. Its core advantage lies in the realization of intelligent management of dynamic demands for complex production materials without increasing the burden on hardware resources by introducing lightweight machine learning mechanisms and traceable dynamic adjustment strategies.

[0077] Furthermore, the update module is specifically used for: Based on the received material status data, combined with historical point material records and production scheduling information, construct a time series feature vector, and by introducing a time-weighted decreasing factor to assign lower weights to early records, obtain a weighted historical sample set for training; Construct a material demand prediction model based on the weighted historical sample set, and dynamically generate a shortage risk level determination threshold according to the volatility label or usage stability grouping of each material number in the material data model to achieve adaptive shortage warning judgment; Based on the comparison between the result output by the prediction model and the current work order data, dynamically adjust the material allocation plan, and generate a dynamic report containing recommended adjustment items, changes in shortage levels, and estimated replenishment windows; Number the generation results of each dynamic report, record the front and back change paths in the storage module, and generate change node records through a graph structure. The nodes include the corresponding material numbers, changes in status fields, prediction differences, and adjustment reasons, which are used to realize version evolution tracking and comparison of the entire process of the dynamic report.

[0078] The update module first receives the material status data provided by the tracking module. This status data includes fields such as the current inventory quantity, consumed quantity, shortage risk level, and locked status of each material number. The system synchronously calls from the historical database the consumed quantity, usage time, error correction information recorded during the past inventory counting operations for this material, as well as the corresponding production scheduling data, such as production batch time, material usage order, production line conversion points, and other information. To enhance the model's perception ability of the time evolution trend, the system constructs a time series feature vector for each material number in chronological order. This vector is used to describe the dynamic evolution of the material status at multiple historical time nodes.

[0079] For example, for a certain material number M1002, its time series feature vector can be composed of the key status fields of this material on each day in the past seven days. For example, in the four dimensions of the inventory quantity, daily consumption, shortage risk level, and whether it is in the locked status on the t-th day, the feature vector of this material can be represented as an array sequence in the form of: [(52, 18, 1, 0), (47, 14, 1, 0), (41, 16, 2, 0), (35, 15, 2, 0), (29, 18, 3, 1), (18, 20, 3, 1), (7, 21, 4, 1)], where each quadruple corresponds to a historical time point, and the field values represent the inventory quantity, daily consumption, shortage risk level (the higher the level, the greater the risk), and locked status (0 means unlocked, 1 means locked) in sequence. This time series feature vector will be used as the input for model training or prediction, supporting the system to identify dynamic features such as an increasing trend of decreasing inventory, rising shortage risk, and frequent changes in the locked status, thereby improving the accuracy and response ability of material demand prediction.

[0080] To avoid excessive interference of early data on the current prediction results, the system introduces a time-weighted decay factor when constructing the sample set. This factor can be set as an exponential decay coefficient based on the time difference. The farther the data is from the current time, the lower its weight, thus improving the model's response ability to recent data. This weighted sample set will be used as the training input and passed into a material demand prediction model based on gradient boosting decision trees or long short-term memory networks (LSTM). This model is used to judge the theoretical consumption trend of each material in the future period under the current work order conditions.

[0081] The system further checks the volatility tags of each material in the material data model. This tag is generated based on historical statistics and is used to distinguish the stability of material usage, such as "daily stable type", "intermittent fluctuation type", or "sensitive to line change type". The system dynamically sets the warning threshold for the out-of-stock risk judgment according to the tag information or usage frequency grouping. For example, for materials with greater volatility, the out-of-stock warning value is set more leniently, while for commonly used materials with high stability, a stricter risk threshold is set, thus enhancing the adaptability and fault tolerance of the system.

[0082] After the material requirement forecasting model completes its operation, the system compares its forecasting results item by item with the original material allocation plan set in the current work order. If it is found that there is a risk of out-of-stock for some materials or the allocated quantity significantly deviates from the forecasted demand, the system will automatically adjust the allocation entries and generate a dynamic report. This report details all the material line items recommended for modification, the change in the out-of-stock level of each material (such as from "normal" to "high risk"), the estimated replenishment window time (such as "replenishment required within 72 hours"), and other fields.

[0083] To ensure the traceability and comparability of report versions, each generated dynamic report is automatically assigned a unique version number by the system. At the same time, it is structurally compared with the previous version, and the change path is written into the database. The system uses a graph structure to record each change node. Each node records the current material number, the name of the field where the change occurs, the content of the field change (such as the inventory changing from 30 to 10), the difference between the model forecast value and the original work order value, and the reason label that triggered this adjustment (such as "forecast exceeds the tolerance range of the work order"). Through this structure, users can clearly view the evolution track of the material allocation adjustment in the interaction module, support backtracking queries and multi-version comparative analysis, and effectively improve the transparency and reliability of the forecast adjustment closed-loop.

[0084] Furthermore, the update module is also used for: Based on the results output by the material requirement forecasting model, compare item by item with the actual material consumption values in the historical spot-check records for the same time period, calculate the forecasting error of each material number at each time node, and generate an error sequence; According to the volatility tags or usage stability grouping of each material number in the material data model, conduct an aggregation analysis on the error sequence. If the errors at multiple consecutive time nodes exceed the forecasting error tolerance threshold corresponding to the volatility grouping, trigger a dynamic reallocation of the weight structure of the weighted historical sample set for reconstructing the material requirement forecasting model; Compare the result output by the reconstructed material requirements forecasting model with the current work order data again, dynamically adjust the material allocation plan, and append a correction mark field to the dynamic report. The correction mark field includes the maximum prediction error, the time range for triggering the correction, and the source of the sample fields involved in this correction. During the process of numbering the generation results of each dynamic report and recording the change path before and after in the storage module, incorporate the above correction mark field into the change node record. The change node record includes the corresponding material number, status field changes, prediction differences, adjustment reasons, and prediction error trends, which are used to achieve complete traceability of the prediction model correction behavior involved in the entire process of the dynamic report.

[0085] First, after the update module generates the material requirements forecasting model, it needs to compare the output result of this forecasting model with the historical point material records stored in the system. The historical point material records contain the actual consumption quantities of each material number during specific time periods in the past real production process. When the update module makes the comparison, it needs to match the material numbers and time nodes one by one, and calculate the difference between the predicted value and the actual consumption value according to daily, shift, batch, or system-defined time granularity, so as to calculate the prediction error of each material number at each time node. All error values are combined in chronological order to form an error sequence, which serves as the basic input for the subsequent model evaluation and correction mechanism.

[0086] After obtaining the error sequence, the update module will retrieve the "volatility label" or "usage stability grouping" of the material number recorded in the material data model. The volatility label is used to identify the usage fluctuation frequency of this material in the past production plan, often classified as high, medium, or low; the usage stability grouping makes a classification setting for its error tolerance ability according to the usage type of this material in the work order, such as "critical raw material", "conventional auxiliary material", or "spare parts", etc. The system automatically loads the corresponding "prediction error tolerance threshold" for this type of material according to the group it belongs to. This threshold may be a fixed numerical range (such as ±10%) or a dynamic proportional function (such as volatility multiplied by coefficient α). If the prediction error exceeds this tolerance threshold at consecutive multiple time nodes (for example, 3 consecutive day shifts or 5 batches), the system determines that the output of the current prediction model for this material number does not have sufficient accuracy.

[0087] When the error over - standard condition is met, the system triggers the model correction mechanism. First, it traces back to the "weighted historical sample set" used in the original training phase. This sample set is composed of historical point - material records and production scheduling information, and is attached with a time - weighted decreasing factor. During the correction process, the system will re - analyze the error sources, identify the most sensitive feature fields affecting the prediction accuracy, such as inventory replenishment fluctuations or production plan mutation points within the time window, etc., and dynamically adjust the weights of these fields in the weighted sample set. The change in the weighted structure will cause a shift in the focus of the model, thus constructing a new material requirement prediction model. This model is different from the original model in terms of parameters, and the optimization goal is more focused on the prediction accuracy of the current cycle.

[0088] After the new prediction model is generated, the update module will compare its output result with the current work - order data again and update the material allocation plan accordingly. At the same time, when the system generates a new dynamic report, it will append a "correction mark field" on the basis of the original field structure. This field is not just a status identifier, but a structured record containing the specific correction background, including: the maximum value of the current prediction error (absolute value or percentage), the continuous time range triggering the correction (such as "April 10, 2025 – April 15, 2025"), and the source of the sample fields re - weighted or excluded during the model correction process (such as "the inventory fluctuation field is increased to a weight of 0.8" or "low - confidence samples are excluded").

[0089] Finally, to ensure the traceability of all report correction processes, when the system executes version numbering for the results of each dynamic report generation, it writes the above "correction mark field" into the storage module. This module will construct the "change path" between the previous and current versions and present it in a graph - structured form, recording the evolution trajectory between nodes. When constructing the change - node record, in addition to basic fields such as "material number", "status field change", "prediction difference", "adjustment reason", content related to prediction correction needs to be added, including "error sequence trend analysis", "model correction trigger flag", "corrected model version number", and "feature fields based on which the correction is made", etc., to form a complete correction behavior chain. The graph structure supports visual rendering of differences between nodes and multi - dimensional comparison, providing users with a complete basis for judging the background, logic, and influence scope of model correction.

[0090] The interaction module 104 is used for data interaction with the acquisition module, tracking module, and update module, and displays the current material data model, point - material records, and dynamic reports; it supports importing the original bill of materials and historical point - material tables simultaneously, completes the classification and mapping of unstructured data based on the field recognition mechanism, and has the functions of resuming point - material interruption and printing dynamic reports and outputting labels based on user permissions.

[0091] The interaction module 104 serves as an information bridge between the system and the user in the present invention. Its implementation process involves not only the front-end display logic but also data calls, feedback flows, and permission control mechanisms among various modules, and must possess complete data reading and writing, format recognition, status recovery, and output control capabilities.

[0092] After the system initialization is completed, the interaction module first establishes a data reading channel with the acquisition module 101, extracts core fields from the generated material data model for the display of basic interface information, including material number, planned quantity, current inventory, supply status, etc. These fields are automatically mapped to column items in the user interface, and the system displays them in the main view in the form of a list or card through front-end dynamic components (such as table components implemented by Vue or React frameworks), and provides filtering, sorting, and multi-dimensional search functions, enabling users to quickly locate specific materials or work orders.

[0093] The interaction module also establishes real-time data linkage with the tracking module 102. The system will dynamically refresh the status fields of the corresponding materials in the interface according to the scanning event or status update request, and use color coding to indicate their current status. For example, green indicates sufficient inventory, yellow indicates approaching shortage, and red indicates out of stock. At the same time, the system supports displaying the material flow path in the form of a visual chart (such as a flow chart, Gantt chart, or timeline chart) on the right side of the interface, along with metadata such as operation records, update time, and operation source, ensuring that users can intuitively understand the dynamic change process of materials.

[0094] To adapt to the situation where there are both structured and unstructured data sources in actual business, the interaction module supports users to import the original material list (such as a BOM table in Excel format) and historical spot-check tables. After the system imports them, it automatically determines the meaning of the header fields through a field recognition mechanism. For example, it uniformly recognizes fields such as "Item Code" and "Part#" as "material number" through regular expression and keyword matching, and prompts the user to confirm the field mapping relationship. If a certain field cannot be automatically recognized, the system provides a field drop-down selection and example preview window for the user to manually specify, so as to uniformly classify the table data into the standard data model structure.

[0095] The interaction module also has a mechanism for resuming interrupted spot-check operations. The system will save the current material data model and the status of the spot-check record to the local cache or distributed database in real time every time the user completes a scanning spot-check or operation submission. When the user fails to complete the operation due to power failure, program shutdown, or human interruption, the system will automatically prompt to resume the previous task when it starts next time, and restore the operation interface to the state before the interruption, including the scanned materials, unfinished items, and historical modification records, ensuring that the task can be carried out continuously.

[0096] In terms of output functions, the interaction module integrates printing and label generation components, which can generate labels with customer information, material numbers, quantities, dates, and barcodes based on the currently selected material records. Users can choose whether to enable 1D barcodes, 2D barcodes, or multi-segment coding modes in the operation interface and can configure label format templates. The system controls whether to allow printing operations according to the user permission settings. For example, ordinary operators can only view and export PDFs, while users with administrative permissions can directly execute batch printing and label generation tasks.

[0097] In addition, to ensure data security and orderly management, the interaction module sets different operation permissions according to user roles. The system identifies user identities through an authentication mechanism (such as username and password or hardware token) and dynamically controls the visibility and editable range of the function menu based on user roles, thus achieving hierarchical management and data protection.

[0098] In summary, the interaction module 104 not only completes data display and import functions, but also realizes the closed-loop data transfer between modules in the system and efficient operations at the user level through mechanisms such as status awareness, field recognition, interruption recovery, and permission output, ensuring the coherence, reliability, and operability of the entire work order material management process.

[0099] Furthermore, the interaction module is specifically used for: During the process of importing the original material list or historical inventory-taking form, perform word vector representation on the table field titles based on natural language processing technology and compare them with the preset standard field set in the system by cosine similarity to achieve automatic classification mapping of unstructured BOM table fields and prompt the field correspondence confidence for users to confirm; When the inventory-taking task is interrupted or the user actively aborts the operation, automatically save the current inventory-taking record status and provide multi-level rollback strategies including task-level rollback, form-level rollback, and field value rollback. The multi-level rollback strategy provides corresponding recovery granularity options based on the preset field importance level and data risk level in the system; When a major change event is triggered in the dynamic report or material status record, the system forcibly updates the corresponding barcode label information, automatically generates a new version label containing the current status summary information, and marks the original label as invalid; among them, the major change events include the transition of the stock-out risk level, the change of the locked state, and the adjustment of the work order instructions; Bind and record the label change information with the dynamic report version number and provide a label history backtracking function in the interaction module to prevent disconnection or mis-matching between the label and the material status, ensuring the consistency and traceability reliability of on-site logistics identification and data status.

[0100] The interaction module is first used to automatically identify and classify the field headers when the user imports the original bill of materials (such as a BOM table in EXCEL format) or the historical stocktaking table. In practical applications, there are significant differences in the field names in the tables exported by different customers, suppliers, or systems, and different spellings such as material name, part name, name, etc. may appear. To solve this problem, the system extracts the text of each field header based on natural language processing technology and converts it into a multi-dimensional word vector representation. The word vector training model can be jointly optimized based on general corpus and industry BOM data. Subsequently, the system calculates the cosine similarity between each word vector and the vectors of all fields in the preset standard field set one by one. The calculation results reflect the semantic proximity between each imported field and the standard field. The system displays the closest standard field and its similarity confidence value on the interface for the user to confirm item by item or manually adjust the mapping relationship during the import process, thus realizing the automatic classification of unstructured fields.

[0101] During the stocktaking operation, the system supports the interruption recovery function. When the user fails to complete the current stocktaking task due to operation interruption, system shutdown, or other abnormal events, the interaction module automatically saves the current stocktaking status to the local cache or the server database. To meet different granularity recovery requirements, the system designs a multi-level rollback strategy. Task-level rollback means restoring to the initial state at the start of the entire stocktaking task, which is applicable to scenarios where the task is determined to be invalid as a whole or needs to be redone entirely; form-level rollback allows restoring to the state point when the previous form was submitted, which is applicable to quickly rolling back when there are mistakes in partial data operations; field value rollback provides a more refined recovery ability, allowing the user to restore the historical values of individual fields item by item. The system dynamically evaluates during operation based on the importance levels of each field (such as "critical material number", "consumption quantity") and the data risk levels (such as "manually input field", "system predicted field") and prompts the most suitable recovery granularity suggestion in the rollback interface.

[0102] When there is a major change event in the dynamic report or material status, the system actively triggers the label status synchronization mechanism. Major change events include, but are not limited to: the shortage risk level of a certain material jumps from "low" to "high", a certain material is marked as "locked state" by the system, or a material number is deleted or replaced in the original work order instruction. When such an event occurs, the interaction module will call the label generation logic to regenerate the barcode label based on the latest material status and add dynamic fields such as "status summary", "change time", etc. to the label. At the same time, the previous version of the label is marked as invalid to prevent logistics personnel or on-site operators from misusing the old label. The label version change information will be automatically bound to the corresponding dynamic report version number to ensure that the label and the report status are consistent.

[0103] In addition, the system provides a label history backtracking function in the interactive module interface. At any point in time, the user can select a certain material number to view the evolution record of its label versions, including information such as the generation time of each version, the type of triggering event, the change in status summary, and the label printing record. This mechanism can effectively avoid inconsistencies between the material status and physical labels caused by update delays or misoperations, ensuring the traceability, standardization, and risk controllability of the on-site production logistics operation process.

[0104] Furthermore, the interactive module is also used for: During the process of importing the original bill of materials or historical spot-checking form, the system performs word vector representation on the form field titles based on natural language processing technology, and after comparing the cosine similarity with the preset standard field set of the system, attaches the corresponding field correspondence confidence value to the classification result of each field. The field correspondence confidence value is used to prompt the user to confirm the credibility level of the current classification mapping. When the user confirms or corrects the automatic classification mapping result of the unstructured BOM form fields based on the field correspondence confidence value, the system uses the finally confirmed field classification result as the field source information, binds it to the material status fields involved in the subsequent dynamic report, and writes this field source information into the label metadata of the barcode label during the label generation process to form a field source binding relationship. When major change events such as the risk level of material shortage, locked state change, or work order instruction adjustment are triggered in the dynamic report or material status record, before the system performs the forced update operation on the corresponding barcode label information, it first compares the field source information bound to the label with the field classification result of the fields corresponding to the current dynamic report. When it is found that the field classification result has changed, the system marks the label status as the invalid status and automatically generates a new version label containing the current status summary information. The interactive module is further used to bind and record the field source binding information of the original label with the dynamic report version number, and provide a label history backtracking function, enabling the user to track, verify, and correct the field source consistency between the label and the material status based on the field classification change path, thereby ensuring the consistency and traceability reliability of the on-site logistics identification and data status.

[0105] In the work order material intelligent tracking and dynamic update system based on multi-source data fusion described in the present invention, the interaction module not only undertakes the functions of conventional data display and user operation interface, but also undertakes key tasks such as semantic parsing of material data structure, management of label binding relationships, and tracing of status evolution. To effectively support the import operation of unstructured original bill of materials or historical spot-checking tables, the system first processes the field titles in the user-imported table through natural language processing technology. This processing process uses mainstream word embedding methods, such as Word2Vec, BERT, etc., to convert each field title into a semantic vector representation. Subsequently, the system obtains the vector representation of the corresponding field from its pre-set standard field set and calculates the cosine similarity with the imported field item by item. In this way, the system can calculate the semantic similarity degree between each table field and the standard field, and infer the standard field category that the field is most likely to map to accordingly.

[0106] To improve the transparency and accuracy of automatic classification, the system presents the similarity result of each classification operation as a "field correspondence confidence value" to the user during each field classification process. This confidence value represents the credibility level of the current classification result, facilitating the user to quickly identify fields that may be ambiguous. The system allows the user to confirm the classification results one by one. If the user believes that the system classification is incorrect, the user can manually re-specify the field classification. Finally, all the confirmed or corrected field classification results will be regarded as "field source information" and established a binding relationship with the field mapping structure in the material data model.

[0107] This field source information is not only used for the structured processing of the data model, but also will establish a strong binding relationship with the material status field during the subsequent dynamic report generation process. More importantly, in the barcode label generation link, the system writes the above field source information into the label content as an important part of the label metadata, thus forming a complete mapping relationship chain among the material status, field source, and label. This mechanism ensures that even when the structure or naming method of the original data table changes, the system can automatically identify whether the label status is still valid based on the change of the field source.

[0108] When dynamic reports or material status records trigger major change events, such as out-of-stock risk level transitions, lock status changes, or work order instruction adjustments, etc., the system first attempts to update the original barcode label according to the report generation logic. However, before performing a forced update, the system will actively compare the field classification results of the fields corresponding to the current dynamic report with the field source information recorded in the label metadata. If the system detects that the field source has changed, that is, the current field classification result is inconsistent with the original label binding information, it is determined that the label is no longer valid and should be invalidated. At this time, the system will automatically generate a new label version, update the label information to reflect the current status, and at the same time set the "field source mismatch invalidation" flag status for the original label to prevent it from being used continuously in subsequent logistics or inventory processes.

[0109] To enhance the system's traceability and operation transparency, the interaction module also provides a label history backtracking function. Specifically, the system establishes a binding record between the field source binding information of the original label and the version number of the dynamic report that generated the label, and stores it in the version change record database. Users can call the backtracking interface through the interaction module to view the change path of the field source bound to any label record in the historical version, and judge the cause and process of the label status change, so as to achieve the verification and control of the entire process of label status evolution.

[0110] Through the above complete mechanism, the present invention not only significantly improves the structure mapping accuracy of unstructured material data, but also constructs a label life cycle control mechanism driven by field classification, which has extremely high practicality in complex manufacturing scenarios with frequent material changes or diverse field naming, and can effectively avoid potential problems such as label mismatches and logistics chaos caused by field classification failures, ensuring the data reliability and label consistency of system operation.

[0111] The storage module 105 is used to store the material data model, spot check records, and dynamic reports in a distributed database, and ensure cross-system data consistency and immutability through the blockchain synchronization mechanism.

[0112] The storage module 105 undertakes the core functions of global data persistence and cross-system synchronization in the present invention. Its design must not only meet the efficient reading and writing requirements of structured data, but also ensure the integrity, consistency, and security of data in a multi-terminal and multi-system environment. To achieve the above goals, the module adopts a combined architecture of a distributed database structure and a blockchain synchronization mechanism, and the following details the specific technical implementation process.

[0113] During the operation of the system, the acquisition module 101 first generates a standardized material data model, which includes fields such as basic information of various materials, source identifiers, planned quantities, etc. This data structure is passed into the storage module 105 in the form of a structured form or object. The module writes the data into a distributed database by setting primary key indexes (such as Material_ID, Order_ID, etc.). The database used can be a data management system that supports high concurrency and cluster deployment, such as MongoDB or PostgreSQL. The system administrator can preset the sharding key in advance to support horizontal expansion of large-scale data. For example, for applications across multiple workshops or factories, data can be partitioned by factory number, with each sub-node storing independently, but coordinated access by the global controller.

[0114] In addition to storing the initial data model, the storage module also needs to continuously receive material status changes from the tracking module 102, allocation adjustment results from the update module 103, and spot-check records and label configurations generated in the interaction module 104. To avoid performance bottlenecks caused by frequent reading and writing, the module is configured with a data writing buffer mechanism, which batches the data into the database through an in-memory queue or a write merging strategy, while ensuring that critical status fields such as inventory changes and shortage level adjustments have writing priorities and trigger immediate synchronization.

[0115] In scenarios that support multi-user or multi-terminal collaboration, data consistency becomes particularly critical. For this reason, the storage module 105 introduces a synchronization mechanism based on blockchain technology. In specific implementations, the system constructs a private chain network through a lightweight consortium chain framework (such as Hyperledger Fabric or Fisco BCOS). Every time there is a material status change, report generation, or label output, it is packaged into a data block through a predefined transaction format. The block contains information such as the original data, changed fields, timestamp, operator ID, and signature hash. All blocks are arranged in order on the chain and are confirmed and written through a consensus mechanism (such as PBFT). In this way, the system can ensure that the data held by each terminal node is logically strictly consistent, and any change operation has traceability and immutability.

[0116] In addition, the module also sets up a data version control mechanism. When the original work order is updated or there is a replenishment adjustment, the system automatically retains the snapshot version before the update and establishes a version linked list structure in the database, enabling users to query or restore any historical state at any time. This provides important support for application scenarios such as handling misfilled records, tracing material paths, and auditing report results.

[0117] The storage module also has the function of data access security control. The system restricts the access scope of operators at all levels to the database by setting user role permissions. For example, general operators can only read the material data and inventory counting records related to this work order, while administrators have the permissions to query across work orders and export all data. All accesses need to pass through the API gateway and call the unified authentication interface for token verification to prevent data leakage or unauthorized operations.

[0118] In summary, the storage module 105 combines a high-performance distributed database and a trusted blockchain synchronization solution in terms of structure, and constructs a complete data receiving, processing, confirmation, version management, and security access process logically. It can stably support the system to write multi-source data, synchronize state changes, and ensure persistence in complex business scenarios, ensuring the integrity, security, and high availability of data throughout the entire life cycle.

[0119] Although this application is disclosed above with preferred embodiments, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the protection scope of this application should be subject to the scope defined by the claims of this application.

Claims

1. A work order material intelligent tracking and dynamic update system based on multi-source data fusion, characterized in that: include: The collection module is used to collect work order data from the bill of materials, production schedule, inventory records and supplier database, and build a unified material data model through standardized protocols and semantic algorithms; A tracking module for real-time tracking of material status based on the material data model, including inventory, consumption and shortage risk, and matching barcodes with material numbers through an integrated barcode recognition and ambiguity resolution unit to achieve automatic status update and graphical path recording; The update module receives material status data, combines historical records and usage frequency, predicts material demand based on machine learning algorithms, and adaptively adjusts material allocation to generate dynamic reports associated with the original work order; An interactive module, used to interact with the acquisition module, tracking module and update module data, and display the current material data model, material point record and dynamic report; Supports simultaneous import of original bill of materials and historical material counting tables, completes classification and mapping of unstructured data based on field recognition mechanism, and has the functions of material counting interruption recovery and dynamic report printing and label output based on user permissions; The storage module is used to store the material data model, material counting records and dynamic reports in a distributed database, and ensure the consistency and non-tamperability of data across systems through a blockchain synchronization mechanism.

2. The intelligent tracking and dynamic updating system for work order materials based on multi-source data fusion according to claim 1 is characterized in that: The tracking module is specifically used for: Based on the barcode scanning event, the current barcode scanned by the user is collected, and the current barcode scanning sequence is constructed in combination with the previous barcode scanning record, and the combined features including the barcode arrangement order, time interval and historical matching accuracy are extracted; According to the combined features of the barcode scanning sequence and the current material status of multiple material numbers that the barcode may correspond to in the material data model, the material status critical value field is identified, and the target material number that is most likely to match the barcode is determined from the multiple material numbers through a context association selection mechanism; Taking the target material number as input, driving the update operation of the material status, marking the matching decision path in the update record, and recording the state evolution path of the material in the material counting process as time series data; Based on the time series data, a multi-dimensional state diagram including state switching frequency, interruption location, and user intervention records is constructed and displayed in the form of a heat map in the system interface for users to predict material shortage risks and make reference for decision-making.

3. The intelligent tracking and dynamic updating system for work order materials based on multi-source data fusion according to claim 2 is characterized in that: The tracking module is also used to: Based on the barcode scanning event, the current barcode scanned by the user is collected, and after the current barcode scanning sequence is constructed in combination with the previous barcode scanning record, the combined features including the barcode arrangement order, time interval and historical matching accuracy are extracted, and further it is determined whether there is an abnormal behavior pattern in the barcode scanning sequence in which the time interval fluctuation exceeds the system preset threshold or the barcode arrangement order deviates from the production process logic model; When it is determined that the current scan sequence contains an abnormal behavior pattern, the abnormal behavior mark is introduced as an input factor into the context association selection mechanism to adjust the weight distribution between multiple material numbers, thereby improving the confidence influence of candidate material numbers with high historical matching accuracy and stable scan context in the determination of the target material number; Using the target material number determined by the contextual association selection mechanism as input, before driving the update operation of the material status, a trusted control strategy for updating the status is added based on the abnormal behavior mark, specifically including marking the match as a pending verification state, limiting the update scope of the material status, and formally submitting it to the material data model after receiving a subsequent completion scan event or manual confirmation by the user; While recording the state evolution path of the material during the counting process as time series data, the path nodes where abnormal behavior intervenes are recorded with obvious marks in the multidimensional state diagram. The marks include path color changes or risk mark superpositions, and are displayed in the form of a heat map in the system interface for users to make predictions on material shortage risks and make reference to error correction decisions based on scanning exceptions.

4. The intelligent tracking and dynamic updating system for work order materials based on multi-source data fusion according to claim 1 is characterized in that: The update module is specifically used for: Based on the received material status data, combined with historical material point records and production scheduling information, a time series feature vector is constructed, and a time-weighted decreasing factor is introduced to give lower weights to early records, thus obtaining a weighted historical sample set for training; A material demand forecasting model is constructed based on the weighted historical sample set, and a material shortage risk level determination threshold is dynamically generated based on the volatility label of each material number in the material data model or the use stability grouping, so as to realize adaptive material shortage early warning judgment; Based on the comparison between the output of the prediction model and the current work order data, the material allocation plan is dynamically adjusted, and a dynamic report including suggested adjustment items, changes in material shortage levels, and estimated replenishment windows is generated; The generation result of each dynamic report is version numbered, and the change path before and after is recorded in the storage module. The change node record is generated through a graphical structure. The node contains the corresponding material number, status field change, predicted difference and adjustment reason, which is used to realize version evolution tracking and comparison of the entire dynamic report process.

5. The intelligent tracking and dynamic updating system for work order materials based on multi-source data fusion according to claim 4 is characterized in that: The update module is also used for: Compare the results output by the material demand forecasting model with the actual material consumption values ​​in the same time period in the historical material point records item by item, calculate the forecast error of each material number at each time node, and generate an error sequence; According to the volatility label of each material number in the material data model or using stability grouping, the error sequence is aggregated and analyzed. If the errors of multiple consecutive time nodes exceed the prediction error tolerance threshold corresponding to the volatility grouping, the weight structure of the weighted historical sample set is dynamically redistributed to rebuild the material demand forecasting model; The output result of the reconstructed material demand forecasting model is compared with the current work order data again, the material allocation plan is dynamically adjusted, and a correction mark field is added to the dynamic report, the correction mark field includes the maximum value of the forecast error, the time range for triggering the correction, and the source of the sample field involved in this correction; In the process of version numbering the generation results of each dynamic report and recording the change paths before and after in the storage module, the above-mentioned correction mark field is included in the change node record. The change node record contains the corresponding material number, status field change, prediction difference, adjustment reason and prediction error trend, which is used to achieve complete traceability of the prediction model correction behavior involved in the entire process of dynamic reporting.

6. The intelligent tracking and dynamic updating system for work order materials based on multi-source data fusion according to claim 1 is characterized in that: The interaction module is specifically used for: When importing the original bill of materials or historical material counting tables, the table field titles are represented by word vectors based on natural language processing technology, and the cosine similarity is compared with the system preset standard field set to achieve automatic classification mapping of unstructured BOM table fields and prompt the field corresponding confidence for user confirmation; When the material counting task is interrupted or the user actively terminates the operation, the current material counting record status is automatically saved, and a multi-level rollback strategy including task-level rollback, form-level rollback and field value rollback is provided. The multi-level rollback strategy provides corresponding recovery granularity options based on the field importance level and data risk level preset in the system; When a major change event is triggered by a dynamic report or material status record, the system will forcibly update the corresponding barcode label information, automatically generate a new version of the label containing the current status summary information, and mark the original label as invalid; wherein, the major change event includes the transition of the risk level of shortage, the change of the lock status, and the adjustment of the work order instruction; The label change information is bound to the dynamic report version number and recorded, and the label history backtracking function is provided in the interactive module to prevent disconnection or mismatch between the label and the material status, and ensure the consistency and tracking reliability of the on-site logistics identification and data status.

7. The intelligent tracking and dynamic updating system for work order materials based on multi-source data fusion according to claim 6 is characterized in that: The interaction module is also used for: When importing the original bill of materials or historical material ordering table, the system uses natural language processing technology to represent the table field titles with word vectors, and compares the cosine similarity with the system's preset standard field set. Then, the corresponding field corresponding confidence value is added to each field classification result. The field corresponding confidence value is used to prompt the user to confirm the trust level of the current classification mapping; When the user confirms or modifies the automatic classification mapping result of the unstructured BOM table field based on the corresponding confidence value of the field, the system uses the finally confirmed field classification result as the field source information, binds it with the material status field involved in the subsequent dynamic report, and writes the field source information into the label metadata of the barcode label during the label generation process to form a field source binding relationship; When a dynamic report or material status record triggers a major change event such as a material shortage risk level transition, a lock status change, or a work order instruction adjustment, the system first performs a consistency comparison between the field source information bound to the label and the field classification result of the field corresponding to the current dynamic report before executing a mandatory update operation on the corresponding barcode label information. When it is found that the field classification result has changed, the system marks the label status as invalid and automatically generates a new version of the label containing the current status summary information; The interactive module is further used to bind and record the field source binding information of the original label with the dynamic report version number, and provide a label history backtracking function, so that the user can track, verify and correct the field source consistency between the label and the material status based on the field classification change path, thereby ensuring the consistency and tracking reliability of the on-site logistics identification and data status.

Images