A warehouse material inventory management system

Abstract

The application provides a workshop material warehouse inventory management system, comprising a real-time sensing module, a data aggregation module, a flow prediction module, a dynamic threshold calculation module and an early warning triggering module. The system collects material inventory data and batch identification data of key process nodes through the real-time sensing module; the data aggregation module standardizes multi-source data and constructs a real-time inventory view; the flow prediction module predicts material transfer timing based on real-time inventory data and generates receiving preparation instructions; the dynamic threshold calculation module calculates dynamic early warning thresholds based on historical consumption data and production plans; and the early warning triggering module generates inventory early warning signals by comparing real-time inventory data with dynamic thresholds. The application solves the problem that existing systems cannot dynamically adjust inventory early warning and production scheduling according to real-time material state and consumption data, realizes dynamic optimization of inventory management and intelligent prediction of production scheduling, and improves the overall efficiency of the production system.

Description

Technical Field

[0001] This invention relates to the field of inventory management technology, and in particular to a workshop material inventory management system. Background Technology

[0002] In the production of condiments like vinegar made from fermented grains, which involve multiple raw materials and fermentation stages, the accuracy of workshop material inventory management plays a decisive role in production efficiency, cost control, and product quality traceability. However, existing workshop material inventory management methods mostly rely on manual recording and periodic inventory checks, failing to provide dynamic tracking and early warning based on real-time consumption and status changes of materials. During the flow of materials in the workshop, their inventory quantity, storage location, and quality status are constantly changing. For example, parameters such as raw material rice, auxiliary material garlic, and in-process mash are dynamically updated with requisition, feeding, fermentation, and transfer. If outdated, static data is consistently used when materials are at different production stages and storage conditions, a series of problems can easily arise.

[0003] In some traditional inventory management systems, although there are some basic inventory warning mechanisms, such as setting a minimum inventory level, these mechanisms can only trigger alarms when the material inventory is lower than the set extreme threshold. They cannot be optimized and adjusted according to the real-time demand of the production plan and the dynamic consumption of materials throughout the entire material flow process. This leads to a disconnect between production planning and execution, inefficient resource allocation, and ultimately lowers the overall efficiency of the entire production system. Summary of the Invention

[0004] In view of this, the present invention proposes a workshop material inventory management system, which solves the technical problem that the existing workshop material inventory management system cannot dynamically adjust inventory warnings and production scheduling based on the real-time status and consumption data of materials, resulting in low overall efficiency of the production system.

[0005] The present invention provides a workshop material inventory management system, comprising: a real-time sensing module, a data aggregation module, a flow prediction module, a dynamic threshold calculation module, and an early warning triggering module;

[0006] The real-time sensing module is deployed at multiple key process nodes in the workshop to collect material status data in real time; wherein, the key process nodes include material handling process nodes and material temporary storage process nodes; the status data includes material inventory data and batch identification data;

[0007] The data aggregation module is used to standardize the status data from different key process nodes and use the standardized data to construct a real-time inventory view of the workshop.

[0008] The flow prediction module is used to predict the timing of material transfer to the material storage process node based on the real-time inventory data of the material handling process node, generate corresponding receiving preparation instructions and send them to the operator terminal.

[0009] The dynamic threshold calculation module is used to calculate the dynamic early warning threshold of the material temporary storage process node based on historical consumption data and production plan.

[0010] The warning triggering module is used to compare the real-time inventory data in the real-time inventory view with the corresponding dynamic warning threshold, and generate an inventory warning signal when the real-time inventory data is less than or equal to the corresponding dynamic warning threshold.

[0011] Optionally, the real-time sensing module includes a first sensing unit, a second sensing unit, and a third sensing unit;

[0012] The first sensing unit is installed on the fermenter, which is the material handling process node, and includes a liquid level sensor and a temperature sensor, which are used to collect the inventory data of the material in the fermenter and the fermentation process temperature data, respectively.

[0013] The second sensing unit is installed on the temporary storage tank, which serves as the material temporary storage process node, and includes a weighing sensor for collecting inventory data of the material in the temporary storage tank.

[0014] The third sensing unit, deployed in the key logistics channels of the workshop, includes multiple fixed RFID readers, used to automatically read the RFID tags attached to the material containers to obtain batch identification data during the transfer process of materials being fed into the fermentation tank or transferred to the temporary storage tank.

[0015] The inventory data and process temperature data collected by the first sensing unit and the second sensing unit are associated and stored in the data aggregation module with the batch identification data collected by the third sensing unit.

[0016] Optionally, the data aggregation module includes a standardization processing unit and a view building unit;

[0017] The standardization processing unit is used to unify the format and parse the protocol of status data from different key process nodes to obtain standardized data.

[0018] The view building unit is used to build and dynamically update the real-time inventory view of the workshop using the standardized data.

[0019] Optionally, the standardization processing unit performs the following steps:

[0020] According to the preset protocol dictionary, the raw data of different communication protocols are parsed into intermediate data with unified semantics;

[0021] Key fields are extracted from the intermediate data, and these key fields are bound to the corresponding batch identifiers and process nodes to obtain initial parsed data;

[0022] For the same production batch, calculate the residual between the output material quantity and the input material quantity between adjacent nodes according to the corresponding process node sequence;

[0023] The residuals are compared with the dynamic error band obtained by statistical learning based on historical production data to generate comparison data;

[0024] Based on the comparison data, the initial parsed data is given a corresponding data quality identifier to generate the target parsed data;

[0025] The target parsed data is formatted and serialized to generate standardized data.

[0026] Optionally, the view building unit performs the following steps:

[0027] The standardized data is deserialized to generate material inventory data, batch identification data, and data quality identification.

[0028] Initialize the graphical framework of the real-time inventory view according to the layout and process node configuration of the workshop;

[0029] Based on the batch identification data and the corresponding process node information, the corresponding graphical elements are dynamically created or updated in the graphical framework.

[0030] Bind the material inventory data to the corresponding graphical elements;

[0031] Visual style data is generated by configuring the corresponding visual style according to the data quality identifier that is already bound to the graphical element.

[0032] The visual style data is used for rendering to generate a real-time inventory view of the workshop.

[0033] Optionally, the flow prediction module includes a data feature extraction unit, a dynamic time series prediction unit, a multi-dimensional instruction generation unit, and an instruction closed-loop feedback unit;

[0034] The data feature extraction unit is used to filter outliers and extract trend features from the real-time inventory data of the material processing process node, and to associate and map the trend features with the material attribute data to obtain standardized predictive input features.

[0035] The dynamic timing prediction unit is used to fuse the standardized prediction input features with historical material transfer timing data and the current status data of the material temporary storage process node, and to build a real-time updated prediction model through adaptive learning, and output the accurate transfer timing of the material to the material temporary storage process node.

[0036] The multi-dimensional instruction generation unit is used to construct a receiving pre-instruction and send it to the operator terminal based on the precise transfer timing, combined with the current task load of the operator terminal, the spatial layout parameters of the material temporary storage process node, and the material storage priority.

[0037] The instruction closed-loop feedback unit is used to collect the instruction confirmation status and execution progress data of the operator terminal in real time. If an execution deviation occurs, it triggers the secondary calibration of the dynamic timing prediction unit and updates the parameters for the next instruction generation.

[0038] Optionally, the instruction closed-loop feedback unit performs the following steps:

[0039] The system collects real-time data on the instruction confirmation status and the actual progress of material transfer returned by the operator terminal.

[0040] The actual execution progress data is compared with the precise transfer timing, and the timing deviation is calculated.

[0041] When the time series deviation exceeds a preset deviation threshold, the internal parameters of the prediction model in the dynamic time series prediction unit are adjusted based on the time series deviation using an adaptive learning algorithm to generate a new accurate transfer time series.

[0042] The new precise transfer timing is used to update the next instruction generation parameters of the multi-dimensional instruction generation unit.

[0043] Optionally, the dynamic threshold calculation module performs the following steps:

[0044] The historical consumption data of the material temporary storage process node is decomposed into time series data to generate time series feature data;

[0045] The time-series feature data and the production plan scheduling data are cross-fused to generate enhanced predictive input features;

[0046] The enhanced predictive input features are input into the consumption prediction model, which outputs the consumption prediction value for future periods.

[0047] The predicted consumption value is multiplied by a preset safety stock coefficient to generate a dynamic early warning threshold for the material temporary storage process node.

[0048] Optionally, the early warning triggering module includes a threshold comparison unit, a strategy enhancement unit, and a signal distribution unit;

[0049] The threshold comparison unit is used to compare the real-time inventory data in the real-time inventory view with the corresponding dynamic warning threshold. When the real-time inventory data is less than or equal to the corresponding dynamic warning threshold, a basic warning signal is generated and output.

[0050] The strategy enhancement unit is used to respond to the basic early warning signal, obtain the corresponding data quality identifier from the real-time inventory view, and perform strategy mapping on the basic early warning signal based on the data quality identifier to generate an enhanced early warning instruction.

[0051] The signal distribution unit is used to execute the enhanced early warning command, generate and distribute the corresponding inventory early warning signal to the designated terminal.

[0052] Optionally, the policy enhancement unit performs the following steps:

[0053] Extract the associated material batch identifier from the basic early warning signal, and obtain the corresponding data quality identifier from the real-time inventory view;

[0054] When the data quality identifier is a first preset type, a first preset enhanced early warning instruction is generated to instruct the execution of replenishment.

[0055] When the data quality identifier is a second preset type, query the associated historical process temperature data based on the material batch identifier;

[0056] When there are abnormal records in the historical process temperature data that continuously deviate from the process range, a second preset enhanced early warning instruction is generated to indicate the quality risk inventory.

[0057] When there are no abnormal records of continuous deviation from the process temperature historical data, a third preset enhanced early warning instruction is generated to indicate inventory verification.

[0058] When the data quality identifier is a third preset type, a fourth preset enhanced warning instruction is generated to indicate data anomalies.

[0059] Compared with the prior art, the beneficial effects of the present invention are:

[0060] This invention effectively solves the problem of inventory warnings and production scheduling lags caused by the inability of existing systems to dynamically adjust based on real-time material status and consumption data by constructing a collaborative operation system of a real-time sensing module, a data aggregation module, a flow prediction module, a dynamic threshold calculation module, and an early warning triggering module. This significantly improves the overall efficiency of the production system. First, a distributed real-time sensing module comprehensively collects material inventory and batch identification data from key process nodes, establishing a precise data foundation. Then, the data aggregation module standardizes multi-source heterogeneous data and constructs a real-time inventory view, eliminating information silos. Based on this, the flow prediction module intelligently predicts material transfer timing based on the real-time inventory of processing nodes and generates receiving preparation instructions, enabling proactive production scheduling. Simultaneously, the dynamic threshold calculation module integrates historical consumption patterns and future production plans to calculate dynamic early warning thresholds that adjust according to supply and demand changes, replacing the traditional fixed threshold mode. Finally, the early warning triggering module generates accurate inventory early warning signals by comparing inventory data with the dynamic thresholds in real time.

[0061] Through real-time data sensing and dynamic view construction, the system achieves transparent management of inventory status, providing data support for accurate decision-making. By intelligently predicting material transfer timing, the system transforms passive response into proactive scheduling, significantly reducing waiting time between processes and improving equipment utilization and the continuity of production rhythm. Through dynamic early warning thresholds based on multi-factor calculations, the system achieves precise control of inventory levels, avoiding production interruptions caused by inventory shortages and preventing capital accumulation due to excessive inventory. Ultimately, the entire system, through a data-driven intelligent early warning and scheduling mechanism, constructs a precise collaborative system from material supply to production consumption, effectively reducing the risk of production stoppages due to poor inventory management, optimizing human resource allocation, and thus significantly improving the overall efficiency, flexibility, and reliability of the production system. Attached Figure Description

[0062] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only preferred embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a schematic diagram of a workshop material inventory management system according to the present invention;

[0064] Figure 2 This is a schematic diagram illustrating the execution steps of the real-time sensing module in a workshop material inventory management system according to the present invention.

[0065] Figure 3This is a schematic diagram illustrating the execution steps of the data aggregation module in a workshop material inventory management system according to the present invention.

[0066] Figure 4 This is a schematic diagram illustrating the execution steps of the flow prediction module in a workshop material inventory management system according to the present invention.

[0067] Figure 5 This is a schematic diagram illustrating the execution steps of the early warning triggering module of a workshop material inventory management system according to the present invention. Detailed Implementation

[0068] This invention provides a workshop material inventory management system to solve the technical problem that existing workshop material inventory management systems cannot dynamically adjust inventory warnings and production scheduling based on the real-time status and consumption data of materials, resulting in low overall efficiency of the production system.

[0069] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0070] See Figure 1 The present invention provides a workshop material inventory management system, comprising: a real-time sensing module, a data aggregation module, a flow prediction module, a dynamic threshold calculation module, and an early warning triggering module.

[0071] The real-time sensing module is deployed at multiple key process nodes in the workshop to collect material status data in real time.

[0072] In this embodiment of the invention, the real-time sensing module adopts a distributed deployment scheme, systematically deployed at key nodes in the workshop process flow. The module's deployment strictly follows the material flow path, constructing an IoT sensing network covering the entire process of feeding, processing, and temporary storage. In terms of deployment, the module uses differentiated configurations based on the functional characteristics of each process node: level and temperature sensors are prioritized at processing nodes to monitor key parameters in the production process; high-precision weighing sensors are configured at temporary storage nodes to ensure the accuracy of inventory data; and an RFID reader / writer network is deployed in the logistics channel to achieve full traceability of material flow. Each sensor synchronously collects data via a PROFIBUS-DP fieldbus with a 500ms sampling period, ensuring the real-time nature and accuracy of the data.

[0073] It should be noted that key process nodes include material handling process nodes and material storage process nodes. Specifically, material handling process nodes refer to fermenters and reactors, responsible for the fermentation and reaction processes of the vinegar made from fermented grains; material storage process nodes refer to storage tanks, used to store semi-finished materials awaiting transfer. Status data includes material inventory data and batch identification data. Material inventory data is collected in real time by level sensors and weighing sensors, reflecting the material inventory at each process node. Batch identification data is obtained through RFID readers, establishing a unique identifier for each material batch. All sensors synchronously collect data via PROFIBUS-DP fieldbus with a 500ms sampling period, ensuring data real-time performance and accuracy.

[0074] The data aggregation module is used to standardize status data from different key process nodes and use the standardized data to build a real-time inventory view of the workshop.

[0075] In this embodiment of the invention, the data aggregation module is built on an industrial server and performs protocol parsing and format unification of multi-source data. A standardized processing unit parses raw data from different communication protocols into intermediate data with unified semantics according to a preset protocol dictionary, and a material balance verification algorithm ensures data quality. The view construction unit uses HTML5 Canvas technology to build a real-time inventory view through three layers: a base layer establishes the process node topology based on the actual workshop layout; a data layer establishes dynamic binding between material inventory data and graphical elements; and a presentation layer implements differentiated display based on data quality identifiers. This ensures that the end-to-end latency from data reception to view update is less than 200ms, providing accurate decision support for managers.

[0076] The flow prediction module is used to predict the timing of material transfer to the material storage process node based on real-time inventory data of the material handling process node, generate corresponding receiving preparation instructions and send them to the operator terminal.

[0077] In this embodiment of the invention, the turnover prediction module is built on the TensorFlow deep learning framework. The data feature extraction unit filters outliers and extracts trend features from real-time inventory data. The dynamic time-series prediction unit fuses standardized prediction input features with historical material transfer time-series data, constructs a prediction model using an LSTM neural network, and outputs accurate transfer time sequences. The multi-dimensional instruction generation unit generates executable pre-receive instructions based on the prediction results, combined with the operator terminal's current task load, the spatial layout parameters of the material temporary storage process node, and the material storage priority. The instruction closed-loop feedback unit continuously optimizes the prediction model parameters by collecting execution progress data in real time.

[0078] The dynamic threshold calculation module is used to calculate the dynamic early warning threshold for material temporary storage process nodes based on historical consumption data and production plans.

[0079] In this embodiment of the invention, historical consumption data of material temporary storage process nodes is decomposed into time series to extract time-series feature data. This time-series feature data is then cross-fused with production planning scheduling data to generate enhanced predictive input features. These enhanced predictive input features are then input into a consumption prediction model, outputting predicted consumption values ​​for future periods. Finally, the predicted consumption values ​​are multiplied by a safety stock coefficient to generate a dynamic early warning threshold. Notably, this module automatically performs the calculation task daily, ensuring the accuracy and timeliness of the threshold settings.

[0080] The early warning triggering module is used to compare the real-time inventory data in the real-time inventory view with the corresponding dynamic early warning threshold. When the real-time inventory data is less than or equal to the corresponding dynamic early warning threshold, an inventory early warning signal is generated.

[0081] In this embodiment of the invention, the early warning triggering module includes a threshold comparison unit, a strategy enhancement unit, and a signal distribution unit. The threshold comparison unit scans real-time inventory data at 100ms intervals, comparing the real-time inventory data in the real-time inventory view with the corresponding dynamic early warning threshold. When the real-time inventory data is less than or equal to the corresponding dynamic early warning threshold, a basic early warning signal is generated and output to promptly detect abnormalities. The strategy enhancement unit responds to the basic early warning signal by obtaining the corresponding data quality identifier from the real-time inventory view and performing strategy mapping on the basic early warning signal based on the data quality identifier to generate an enhanced early warning instruction. The signal distribution unit distributes the enhanced early warning instruction to designated terminals through multiple channels to ensure timely delivery and processing of early warning information.

[0082] Further, please refer to Figure 2 The real-time sensing module includes a first sensing unit, a second sensing unit, and a third sensing unit;

[0083] The first sensing unit is deployed on the fermenter, which is a material handling process node. It includes a liquid level sensor and a temperature sensor, which are used to collect data on the inventory of materials in the fermenter and the fermentation process temperature, respectively.

[0084] The second sensing unit is installed on the temporary storage tank, which serves as a material temporary storage process node. It includes a weighing sensor for collecting inventory data of the material in the temporary storage tank.

[0085] The third sensing unit, deployed in the key logistics channels of the workshop, includes multiple fixed RFID readers, which are used to automatically read the RFID tags attached to the material containers to obtain batch identification data during the transfer process of materials being fed into the fermentation tank or transferred to the temporary storage tank.

[0086] The inventory data and process temperature data collected by the first and second sensing units are associated and stored in the data aggregation module with the corresponding batch identification data collected by the third sensing unit.

[0087] It should be noted that the first sensing unit adopts a split sensor deployment scheme, including a liquid level sensor and a temperature sensor. The liquid level sensor uses an E+H FMU95 ultrasonic level gauge, with a measurement range of 0-5 meters, an accuracy of ±2 mm, and outputs a 4-20mA analog signal. The temperature sensor uses a WIKA TR-21 platinum resistance thermometer, with a measurement range of 0-100℃ and an accuracy of ±0.5℃. Both sensors communicate with a Siemens S7-1200 PLC via a PROFIBUS-DP bus, with a sampling period set to 500ms. Considering the characteristics of the fermentation process for vinegar, the system installs sensors on the top and side walls of the fermenter, respectively. A weighted average algorithm is used to calculate the actual liquid level, effectively eliminating the influence of the foam layer on the measurement. This accurately collects data on the inventory of materials in the fermenter and the fermentation process temperature.

[0088] The second sensing unit uses a Mettler Toledo IND560 digital weighing terminal, paired with an HBM PW15C load cell, with a weighing range of 0-10 tons and an accuracy class of C3. The load cell uses a six-wire connection and temperature compensation technology to effectively overcome the impact of ambient temperature changes on weighing accuracy, collecting inventory data of materials in the temporary storage tank. The system also features automatic zero-point tracking, performing automatic calibration every 24 hours to ensure long-term measurement stability.

[0089] The third sensing unit comprises multiple fixed RFID readers deployed at key logistics channels in the workshop, including the feeding port, discharging port, and transfer channel. The RFID readers are SICK RFU650, operating at a frequency of 865-868MHz, with an adjustable reading distance of 0.5-3 meters. The RFID tags are anti-metal ceramic tags, directly affixed to prominent locations on the material containers. During the transfer process of materials from the fermentation tank to the temporary storage tank, the system automatically reads the RFID tags attached to the material containers to obtain batch identification data. Furthermore, the system incorporates an anti-collision algorithm, supports simultaneous reading of multiple tags, and achieves a success rate of ≥99.8%.

[0090] In this embodiment of the invention, the data collected by the three sensing units are correlated through a unified time sequence identifier. When material is transferred from the fermenter to the temporary storage tank, the system records the transfer start timestamp, the third sensing unit reads the material batch identifier, the first sensing unit records the liquid level data before transfer, and the second sensing unit records the weight data after receipt. The data aggregation module establishes a complete material transfer file based on the timestamp and batch identifier, including key information such as transfer time, transfer volume, and process temperature curve. Through the collaborative work of the three sensing units, the system achieves precise monitoring of the entire process of material feeding to discharge, providing complete and accurate basic data support for workshop material inventory management.

[0091] Further, please refer to Figure 3 The data aggregation module includes a standardized processing unit and a view building unit connected in sequence;

[0092] The standardization processing unit is used to unify the format and parse the protocol of status data from different key process nodes to obtain standardized data.

[0093] The view building unit is used to build and dynamically update a real-time inventory view of the shop floor using standardized data.

[0094] In this embodiment of the invention, the standardization processing unit receives status data from different key process nodes and unifies the format of multi-source heterogeneous data through a pre-configured protocol dictionary. This unit supports mainstream industrial protocols such as PROFIBUS-DP, ModbusTCP, and OPC UA, completing protocol parsing through three steps: address mapping, data type conversion, and unit unification, to obtain standardized data. The view construction unit adopts a front-end / back-end separation architecture. The front-end uses the Vue.js framework and ECharts visualization library for data rendering, while the back-end provides data interfaces through Spring Boot microservices. Using standardized data, a real-time inventory view of the workshop is constructed and dynamically updated. The two units establish a real-time data channel via WebSocket, ensuring that the end-to-end latency from data reception to view update is controlled within 200ms.

[0095] It should be noted that the standardization processing unit employs a multi-level processing mechanism to ensure data quality. This mechanism includes: first, protocol parsing to convert raw data into a unified format; second, verifying the logical rationality of the data through a material balance model; and third, adding metadata such as timestamps and quality identifiers to valid data. The view construction unit uses incremental updates and differential rendering techniques, updating only changed data locally and implementing differentiated display strategies based on data quality identifiers. This design significantly improves the real-time performance and user experience of visualization while ensuring data processing accuracy. Through deep collaboration between the two units, the data aggregation module ensures that downstream modules receive high-quality standardized data and provides managers with an intuitive and accurate decision support interface.

[0096] Preferably, the standardized processing unit performs the following steps:

[0097] According to the preset protocol dictionary, the raw data of different communication protocols are parsed into intermediate data with unified semantics;

[0098] Extract key fields from intermediate data and bind them with corresponding batch identifiers and process nodes to obtain initial parsed data;

[0099] For the same production batch, calculate the residual between the output material quantity and the input material quantity between adjacent nodes according to the corresponding process node sequence;

[0100] Comparison data is generated by comparing the residuals with the dynamic error band obtained from statistical learning based on historical production data.

[0101] Based on the comparison data, the initial parsed data is given the corresponding data quality labels to generate the target parsed data;

[0102] The target parsed data is formatted and serialized to generate standardized data.

[0103] In this embodiment of the invention, semantic parsing is performed on raw data from different key process nodes based on a preset protocol dictionary to obtain intermediate data with unified semantics. It should be noted that the preset protocol dictionary refers to a pre-configured set of protocol mapping rules, whose core function is to achieve semantic unification of multi-source heterogeneous data. This dictionary is stored in XML format and contains mapping rules for protocol types, data point addresses, data types, and units of measurement, and supports online dynamic updates. In actual operation, the mapping parameters of the protocol dictionary need to be configured according to the specific communication specifications of the workshop equipment. These parameters typically include Modbus TCP register address mapping, PROFIBUS-DP slot number configuration, and OPC UA node path mapping, ensuring the accuracy and completeness of data parsing.

[0104] Key fields in the data, including material codes, production timestamps, and equipment numbers, are identified using feature matching algorithms. These fields are then bound to batch identifiers and process node information collected by the real-time sensing module to obtain initial parsed data. In actual deployment, standardized field extraction rules need to be established to ensure the accuracy of key field identification, while maintaining a stable binding relationship mapping table.

[0105] For materials in the same production batch, the residual between the output and input material quantities of adjacent nodes is calculated according to the process node sequence. It should be noted that residual calculation refers to a balance check based on the principle of material conservation, and its core purpose is to detect anomalies in the material flow process. The calculation cycle is set to 5 minutes, using a sliding window mechanism. The window size is dynamically adjusted according to the production rhythm, typically set to data from the most recent 10 calculation cycles.

[0106] The calculated residuals are compared with a dynamic error band derived from statistical learning of historical production data. It's important to note that the dynamic error band refers to an adaptive threshold range obtained through statistical learning of historical production data; its core function is to provide a criterion for judging the reasonableness of the data. This error band is calculated using an exponentially weighted moving average algorithm based on the analysis of production data from the past 90 days, and it can automatically adjust the threshold range according to changes in production conditions. During the comparison process, the system records the degree of deviation between the residuals and the error band, generating comparative data that includes the deviation value and direction.

[0107] Based on the comparative data, data quality labels are added to the initial parsed data. These labels are graded tags used to characterize the reliability of the data, employing a three-level classification system: "Reliable" indicates that the data has passed all checks and the residuals are within the error band; "Questionable" indicates that the data has slight deviations but can be manually verified; and "Unusable" indicates that the data has significant anomalies and requires immediate action. In practical applications, operators need to configure the labeling rules according to quality requirements. Typically, residuals within ±5% of the error band are set as "Reliable," ±5%-15% as "Questionable," and exceeding ±15% as "Unusable."

[0108] The target parsed data is converted into a unified, standardized format, specifically using the Apache Avro binary serialization protocol, achieving efficient encoding through predefined data patterns. In actual deployment, serialization parameters need to be configured, including compression algorithms, data block sizes, and checksum mechanisms, to ensure the efficiency and reliability of data transmission. Then, through serialization encapsulation, standardized data is obtained, reducing data volume by an average of 60% while maintaining the complete data structure and metadata.

[0109] Preferably, the view building unit performs the following steps:

[0110] Standardized data is deserialized to generate material inventory data, batch identification data, and data quality identification.

[0111] Initialize the graphical framework of the real-time inventory view according to the layout of the workshop and the configuration of process nodes;

[0112] Based on batch identification data and corresponding process node information, the corresponding graphical elements are dynamically created or updated in the graphical framework.

[0113] Bind material inventory data to corresponding graphical elements;

[0114] Configure the corresponding visual styles according to the data quality identifiers of the bound graphical elements, and generate visual style data;

[0115] Visual style data is used for rendering operations to generate a real-time inventory view of the workshop.

[0116] In this embodiment of the invention, the view construction unit is implemented based on the Web technology stack. This unit deserializes the standardized data serialized by Apache Avro, parses and restores three key structured data items, including material inventory data, batch identification data, and data quality identifiers assigned by upstream modules. Based on the pre-defined digital workshop floor plan and equipment list, i.e., according to the workshop layout and process node configuration, the underlying graphical framework of the real-time inventory view is initialized using SVG vector graphics technology. This framework outlines the location and connection relationships of key process nodes such as fermenters, temporary storage tanks, and logistics channels. Then, based on the parsed batch identification data, corresponding graphical elements are dynamically created or updated at the corresponding process node locations within the graphical framework. For example, when a new batch of materials enters the fermenter, the system dynamically creates a new graphical element representing this batch next to the tank icon.

[0117] Material inventory data is dynamically and continuously linked to graphical elements, ensuring that any changes in backend data are reflected in the frontend display in real time. Based on data quality indicators, specific display attributes are configured for each linked graphical element, and visual style data driving the interface rendering is generated. For example, data identified as reliable has a green border, while data identified as suspicious is displayed in yellow and flashes slowly. The rendering engine uses this visual style data to draw all graphical elements, ultimately outputting a complete, intuitive, and real-time updated view of the workshop's inventory.

[0118] Further, please refer to Figure 4 The flow prediction module includes a data feature extraction unit, a dynamic time series prediction unit, a multi-dimensional instruction generation unit, and an instruction closed-loop feedback unit.

[0119] The data feature extraction unit is used to filter outliers and extract trend features from real-time inventory data of material handling process nodes, and to correlate and map the trend features with material attribute data to obtain standardized predictive input features.

[0120] The dynamic timing prediction unit is used to fuse standardized prediction input features with historical material transfer timing data and current status data of material temporary storage process nodes, and build a real-time updated prediction model through adaptive learning to output accurate transfer timing of material to material temporary storage process nodes.

[0121] The multi-dimensional instruction generation unit is used to construct and send pre-received instructions to the operator terminal based on the precise transfer timing, combined with the current task load of the operator terminal, the spatial layout parameters of the material temporary storage process node, and the material storage priority.

[0122] The instruction closed-loop feedback unit is used to collect instruction confirmation status and execution progress data from the operator terminal in real time. If an execution deviation occurs, it triggers the secondary calibration of the dynamic timing prediction unit and updates the parameters for the next instruction generation.

[0123] In this embodiment of the invention, the data feature extraction unit employs a sliding window mechanism to smooth the real-time inventory data of material handling process nodes, effectively identifying and eliminating abnormal readings caused by equipment vibration or signal interference. By analyzing the slope of the inventory curve, core trend features characterizing the consumption rate are extracted, and these trend features are correlated and mapped with the inherent attribute data of the materials, ultimately generating standardized predictive input features that conform to the model input specifications. The dynamic time-series prediction unit deeply integrates the cleaned and feature-engineered standardized predictive input features with long-accumulated historical material transfer time-series data and real-time status data of material temporary storage process nodes. Based on this multi-dimensional information, the unit uses a long short-term memory neural network as the basic predictive architecture to construct a real-time updated predictive model. Through an online learning mechanism, the model can dynamically adjust its internal weights according to the latest production rhythm, thereby continuously outputting high-precision and accurate transfer time sequences.

[0124] The multi-dimensional instruction generation unit not only considers precise transfer timing but also comprehensively evaluates multiple practical factors such as the current task load of the operator's terminal, spatial layout parameters within the workshop, and the storage priorities of different materials. Through a built-in scheduling algorithm, multi-objective optimization is performed to ultimately generate reasonable and efficient pre-receive instructions, which are then pushed to the designated operator's terminal. The instruction closed-loop feedback unit continuously monitors the entire process from instruction issuance to actual execution, collecting key feedback information including instruction confirmation status and execution progress data. By comparing actual data with the predicted precise transfer timing, when a significant execution deviation is detected, the unit immediately initiates a calibration procedure, triggering the dynamic timing prediction unit to perform secondary calibration to correct model parameters. Simultaneously, the results of this deviation analysis are used to update the next instruction generation parameters, thus forming a continuous improvement closed loop from practical feedback to predictive decision-making, significantly enhancing the system's adaptability and reliability in complex production environments.

[0125] Preferably, the instruction closed-loop feedback unit performs the following steps:

[0126] Real-time acquisition of instruction confirmation status and actual material transfer progress data returned by operator terminals;

[0127] The actual execution progress data is compared with the precise transfer timeline, and the timeline deviation is calculated.

[0128] When the time series deviation exceeds the preset deviation threshold, the internal parameters of the prediction model in the dynamic time series prediction unit are adjusted based on the time series deviation using an adaptive learning algorithm to generate a new accurate transfer time series.

[0129] A new precise transfer timing is used to update the next instruction generation parameters of the multi-dimensional instruction generation unit.

[0130] The preset deviation threshold refers to a critical value that is set in advance to determine whether the deviation between the output of the prediction model and the actual execution is within an acceptable range.

[0131] In this embodiment of the invention, the instruction closed-loop feedback unit first collects instruction confirmation status from operator terminals in real time through the workshop IoT gateway, as well as actual material transfer progress data obtained through the real-time sensing module. These data collectively constitute the objective basis for evaluating the effectiveness of the predicted instruction execution. Based on the data collection, the instruction closed-loop feedback unit precisely compares the recorded actual execution progress data, especially the actual time points of material transfer, with the previously output accurate transfer time sequence from the dynamic time-series prediction unit. By calculating the time difference between the two, a quantified time-series deviation is obtained, which directly reflects the degree of consistency between the prediction accuracy and actual production execution. When the calculated time-series deviation exceeds a preset deviation threshold, the system determines that a significant deviation needs correction. At this time, the instruction closed-loop feedback unit initiates a calibration process. Based on this time-series deviation, it drives an embedded adaptive learning algorithm to fine-tune the internal parameters of the prediction model in the dynamic time-series prediction unit. This adjustment process aims to ensure that the model's output, i.e., the newly generated accurate transfer time sequence, better aligns with the actual production rhythm and operating habits. Finally, the optimized newly generated accurate transfer time sequence is synchronized to the multi-dimensional instruction generation unit. This update operation directly affects the core time reference when the unit builds and receives preparatory instructions next time, thereby ensuring that subsequent production instructions can inherit the latest learning results and achieve iterative improvement in prediction accuracy.

[0132] Furthermore, the dynamic threshold calculation module performs the following steps:

[0133] Perform time-series decomposition on historical consumption data of material temporary storage process nodes to generate time-series feature data;

[0134] By cross-fusing time-series feature data with production planning scheduling data, enhanced predictive input features are generated.

[0135] The enhanced predictive input features are input into the consumption prediction model, which outputs the consumption prediction value for future periods.

[0136] The dynamic early warning threshold for the material temporary storage process node is generated by multiplying the predicted consumption value with the preset safety stock coefficient.

[0137] A consumption prediction model is a mathematical model trained using machine learning algorithms to accurately predict the consumption of materials in a specific future period.

[0138] The preset safety stock factor is a buffer multiplier set in advance to cope with supply and demand uncertainties. Its function is to add an extra safety stock amount based on the predicted consumption.

[0139] In this embodiment of the invention, the dynamic threshold calculation module performs in-depth time-series analysis on the historical consumption data of material temporary storage process nodes. It uses a seasonal trend decomposition method to break down the original data into a combination of trend components, seasonal components, and residual components. These components together constitute time-series characteristic data reflecting the material consumption pattern. The time-series characteristic data is then cross-fused with the scheduling data of the production plan, combining the statistical regularity of the consumption data with the operational requirements of future production arrangements. Through feature engineering, an enhanced predictive input feature that includes both historical patterns and future expectations is constructed. The consumption prediction model is pre-trained based on historical consumption data and production plan data using an ensemble learning algorithm. This model takes the enhanced predictive input feature obtained by fusing the time-series characteristic data and scheduling data as input and outputs a consumption prediction value. In actual deployment, the model is periodically retrained and updated to ensure its prediction accuracy. By multiplying the obtained consumption prediction value with a preset safety stock coefficient, a dynamic early warning threshold that conforms to actual consumption patterns while retaining an appropriate safety margin is finally generated.

[0140] Further, please refer to Figure 5 The early warning triggering module includes a threshold comparison unit, a strategy enhancement unit, and a signal distribution unit;

[0141] The threshold comparison unit is used to compare the real-time inventory data in the real-time inventory view with the corresponding dynamic warning threshold. When the real-time inventory data is less than or equal to the corresponding dynamic warning threshold, a basic warning signal is generated and output.

[0142] The strategy enhancement unit is used to respond to basic early warning signals, obtain the corresponding data quality identifier from the real-time inventory view, and perform strategy mapping on the basic early warning signals based on the data quality identifier to generate enhanced early warning instructions.

[0143] The signal distribution unit is used to execute enhanced early warning instructions, generate and distribute corresponding inventory early warning signals to designated terminals.

[0144] In this embodiment of the invention, the early warning triggering module realizes intelligent generation and distribution of early warning signals through a three-level processing architecture. The early warning triggering module includes a threshold comparison unit, a strategy enhancement unit, and a signal distribution unit. The threshold comparison unit, as the triggering source of the module, continuously scans the real-time inventory data in the real-time inventory view at a period of 100 milliseconds, comparing it in real-time with the dynamic early warning threshold provided by the dynamic threshold calculation module. When it detects that the real-time inventory data of any material temporary storage process node reaches or falls below its corresponding dynamic early warning threshold, the unit immediately generates and outputs a basic early warning signal containing key information such as node location, material batch, and inventory quantity. The strategy enhancement unit, as the intelligent decision-making core of the module, is responsible for in-depth processing of the basic early warning signal. After receiving the basic early warning signal, this unit extracts the corresponding data quality identifier from its associated real-time inventory view. Based on the data credibility level reflected by this identifier, the unit performs intelligent mapping through a built-in strategy rule base, transforming the unified basic early warning signal into an enhanced early warning instruction with a clear handling orientation. The signal distribution unit, as the execution output of the module, is responsible for translating decisions into actions. By analyzing the specific requirements of enhanced early warning instructions, an inventory early warning signal containing detailed operational guidance is generated. Based on the early warning level and type, the signal is distributed to the corresponding designated terminals through multiple channels, including SMS, email, and system pop-ups. This tiered and distributed early warning mechanism ensures timely response to emergencies while avoiding unnecessary production disruptions, thus improving the practicality and reliability of the entire inventory management system.

[0145] Preferably, the policy enhancement unit performs the following steps:

[0146] Extract the associated material batch identifier from the basic early warning signal, and at the same time obtain the corresponding data quality identifier from the real-time inventory view;

[0147] When the data quality is identified as the first preset type, a first preset enhanced warning instruction is generated to indicate the execution of replenishment.

[0148] When the data quality identifier is the second preset type, query the associated historical process temperature data based on the material batch identifier;

[0149] When there are abnormal records in the historical process temperature data that continuously deviate from the process range, a second preset enhanced early warning instruction is generated to indicate the quality risk inventory.

[0150] When there are no abnormal records of continuous deviation from the process temperature historical data, a third preset enhanced early warning instruction is generated to indicate inventory verification.

[0151] When the data quality is identified as the third preset type, a fourth preset enhanced warning instruction is generated to indicate data anomalies.

[0152] It should be noted that the first preset type corresponds to a reliable data state, indicating that the batch of data has passed material balance verification and the residuals are within the normal fluctuation range. The second preset type corresponds to a questionable data state, indicating that the data has deviations exceeding the normal range but still maintains basic integrity. The third preset type corresponds to an abnormal data state, indicating that the data contains serious errors or omissions and has lost its reference value.

[0153] The first preset enhanced early warning instruction is a standard replenishment instruction triggered by reliable data. The second preset enhanced early warning instruction is a quality risk handling instruction triggered when data is questionable and process temperature is abnormal. The third preset enhanced early warning instruction is an inventory check instruction triggered when data is questionable but process parameters are normal. The fourth preset enhanced early warning instruction is an equipment maintenance instruction triggered by abnormal data.

[0154] In this embodiment of the invention, the strategy enhancement unit first parses the key material batch identifier from the received basic warning signal, and uses this as an index to retrieve the corresponding data quality identifier from the real-time inventory view. When the data quality identifier is of a first preset type, it indicates that the relevant inventory data is completely reliable. At this time, the unit directly generates a first preset enhanced warning instruction, which will drive subsequent modules to execute the standard replenishment process.

[0155] When the data quality is identified as the second preset type, it indicates that the data is uncertain. In this case, the unit will further query the historical process temperature data of the material batch during the material handling process node based on the material batch identifier. By analyzing the degree of conformity between the temperature curve and the standard process range, the unit can make an accurate judgment: if an abnormal record with continuous deviation is found, a second preset enhanced warning instruction is generated, suggesting isolation and quality verification of the relevant batch; if the temperature record is normal, a third preset enhanced warning instruction is generated, suggesting priority to manual inventory confirmation.

[0156] When the data quality is identified as the third preset type, it indicates that the data quality is unreliable. The unit then generates a fourth preset enhanced warning command to remind maintenance personnel to check the sensors and data acquisition links.

[0157] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A workshop material inventory management system, characterized in that, include: The system includes a real-time sensing module, a data aggregation module, a flow prediction module, a dynamic threshold calculation module, and an early warning triggering module. The real-time sensing module is deployed at multiple key process nodes in the workshop to collect material status data in real time; wherein, the key process nodes include material handling process nodes and material temporary storage process nodes; the status data includes material inventory data and batch identification data; The data aggregation module is used to standardize the status data from different key process nodes and use the standardized data to construct a real-time inventory view of the workshop. The flow prediction module is used to predict the timing of material transfer to the material storage process node based on the real-time inventory data of the material handling process node, generate corresponding receiving preparation instructions and send them to the operator terminal. The dynamic threshold calculation module is used to calculate the dynamic early warning threshold of the material temporary storage process node based on historical consumption data and production plan. The warning triggering module is used to compare the real-time inventory data in the real-time inventory view with the corresponding dynamic warning threshold, and generate an inventory warning signal when the real-time inventory data is less than or equal to the corresponding dynamic warning threshold.

2. The workshop material inventory management system according to claim 1, characterized in that, The real-time sensing module includes a first sensing unit, a second sensing unit, and a third sensing unit; The first sensing unit is installed on the fermenter, which is the material handling process node, and includes a liquid level sensor and a temperature sensor, which are used to collect the inventory data of the material in the fermenter and the fermentation process temperature data, respectively. The second sensing unit is installed on the temporary storage tank, which serves as the material temporary storage process node, and includes a weighing sensor for collecting inventory data of the material in the temporary storage tank. The third sensing unit, deployed in the key logistics channels of the workshop, includes multiple fixed RFID readers, used to automatically read the RFID tags attached to the material containers to obtain batch identification data during the transfer process of materials being fed into the fermentation tank or transferred to the temporary storage tank. The inventory data and process temperature data collected by the first sensing unit and the second sensing unit are associated and stored in the data aggregation module with the batch identification data collected by the third sensing unit.

3. The workshop material inventory management system according to claim 1, characterized in that, The data aggregation module includes a standardization processing unit and a view construction unit; The standardization processing unit is used to unify the format and parse the protocol of status data from different key process nodes to obtain standardized data. The view building unit is used to build and dynamically update the real-time inventory view of the workshop using the standardized data.

4. A workshop material inventory management system according to claim 3, characterized in that, The standardization processing unit performs the following steps: According to the preset protocol dictionary, the raw data of different communication protocols are parsed into intermediate data with unified semantics; Key fields are extracted from the intermediate data, and these key fields are bound to the corresponding batch identifiers and process nodes to obtain initial parsed data; For the same production batch, calculate the residual between the output material quantity and the input material quantity between adjacent nodes according to the corresponding process node sequence; The residuals are compared with the dynamic error band obtained by statistical learning based on historical production data to generate comparison data; Based on the comparison data, the initial parsed data is given a corresponding data quality identifier to generate the target parsed data; The target parsed data is formatted and serialized to generate standardized data.

5. A workshop material inventory management system according to claim 3, characterized in that, The view building unit performs the following steps: The standardized data is deserialized to generate material inventory data, batch identification data, and data quality identification. Initialize the graphical framework of the real-time inventory view according to the layout and process node configuration of the workshop; Based on the batch identification data and the corresponding process node information, the corresponding graphical elements are dynamically created or updated in the graphical framework. Bind the material inventory data to the corresponding graphical elements; Visual style data is generated by configuring the corresponding visual style according to the data quality identifier that is already bound to the graphical element. The visual style data is used for rendering to generate a real-time inventory view of the workshop.

6. A workshop material inventory management system according to claim 1, characterized in that, The flow prediction module includes a data feature extraction unit, a dynamic time series prediction unit, a multi-dimensional instruction generation unit, and an instruction closed-loop feedback unit. The data feature extraction unit is used to filter outliers and extract trend features from the real-time inventory data of the material processing process node, and to associate and map the trend features with the material attribute data to obtain standardized predictive input features. The dynamic timing prediction unit is used to fuse the standardized prediction input features with historical material transfer timing data and the current status data of the material temporary storage process node, and to build a real-time updated prediction model through adaptive learning, and output the accurate transfer timing of the material to the material temporary storage process node. The multi-dimensional instruction generation unit is used to construct a receiving pre-instruction and send it to the operator terminal based on the precise transfer timing, combined with the current task load of the operator terminal, the spatial layout parameters of the material temporary storage process node, and the material storage priority. The instruction closed-loop feedback unit is used to collect the instruction confirmation status and execution progress data of the operator terminal in real time. If an execution deviation occurs, it triggers the secondary calibration of the dynamic timing prediction unit and updates the parameters for the next instruction generation.

7. A workshop material inventory management system according to claim 6, characterized in that, The instruction closed-loop feedback unit performs the following steps: The system collects real-time data on the instruction confirmation status and the actual progress of material transfer returned by the operator terminal. The actual execution progress data is compared with the precise transfer timing, and the timing deviation is calculated. When the time series deviation exceeds a preset deviation threshold, the internal parameters of the prediction model in the dynamic time series prediction unit are adjusted based on the time series deviation using an adaptive learning algorithm to generate a new accurate transfer time series. The new precise transfer timing is used to update the next instruction generation parameters of the multi-dimensional instruction generation unit.

8. A workshop material inventory management system according to claim 1, characterized in that, The dynamic threshold calculation module performs the following steps: The historical consumption data of the material temporary storage process node is decomposed into time series data to generate time series feature data; The time-series feature data and the production plan scheduling data are cross-fused to generate enhanced predictive input features; The enhanced predictive input features are input into the consumption prediction model, which outputs the consumption prediction value for future periods. The predicted consumption value is multiplied by a preset safety stock coefficient to generate a dynamic early warning threshold for the material temporary storage process node.

9. A workshop material inventory management system according to claim 1, characterized in that, The early warning triggering module includes a threshold comparison unit, a strategy enhancement unit, and a signal distribution unit; The threshold comparison unit is used to compare the real-time inventory data in the real-time inventory view with the corresponding dynamic warning threshold. When the real-time inventory data is less than or equal to the corresponding dynamic warning threshold, a basic warning signal is generated and output. The strategy enhancement unit is used to respond to the basic early warning signal, obtain the corresponding data quality identifier from the real-time inventory view, and perform strategy mapping on the basic early warning signal based on the data quality identifier to generate an enhanced early warning instruction. The signal distribution unit is used to execute the enhanced early warning command, generate and distribute the corresponding inventory early warning signal to the designated terminal.

10. A workshop material inventory management system according to claim 9, characterized in that, The policy enhancement unit performs the following steps: Extract the associated material batch identifier from the basic early warning signal, and obtain the corresponding data quality identifier from the real-time inventory view; When the data quality identifier is a first preset type, a first preset enhanced early warning instruction is generated to instruct the execution of replenishment. When the data quality identifier is a second preset type, query the associated historical process temperature data based on the material batch identifier; When there are abnormal records in the historical process temperature data that continuously deviate from the process range, a second preset enhanced early warning instruction is generated to indicate the quality risk inventory. When there are no abnormal records of continuous deviation from the process temperature historical data, a third preset enhanced early warning instruction is generated to indicate inventory verification. When the data quality identifier is a third preset type, a fourth preset enhanced warning instruction is generated to indicate data anomalies.

Images