AI and RFID based mold full life cycle digital management and predictive maintenance system

Abstract

The application discloses a mold full life cycle digital management and predictive maintenance system based on AI and RFID, relates to the field of big data management, and comprises the following steps: configuring a unique RFID tag for a mold to realize identity binding, integrating multiple source sensors and RFID equipment to collect mold operation data, quality data and use time length in real time, and transmitting the data to the cloud by relying on 5G / industrial internet of things; the cloud stores the data by means of a distributed database, constructs a dynamically updated mold health evaluation AI model based on machine learning and deep learning algorithms, realizes real-time health diagnosis and fault early warning, and generates maintenance suggestions; integrating mold full process data, providing a visual traceability query function, and feeding actual maintenance records back to the AI model to form a closed-loop optimization. The application has the advantages that the mold health state can be accurately evaluated, faults can be early warned, and targeted maintenance suggestions can be generated, full-process traceability and maintenance data closed-loop optimization are completed, mold management efficiency is greatly improved, maintenance cost is reduced, and the service life of the mold is prolonged.

Description

Technical Field

[0001] This invention relates to the field of big data management, and in particular to a mold full lifecycle digital management and predictive maintenance system based on AI and RFID. Background Technology

[0002] As a core piece of equipment in the manufacturing industry, mold management throughout its entire lifecycle directly impacts production efficiency and costs. Traditional mold management relies on manual experience, resulting in problems such as data silos, delayed maintenance, and inaccurate lifespan predictions, leading to frequent unplanned downtime and significant resource waste. The maturity of digital twins, the Internet of Things, and big data analytics technologies provides new approaches to mold management.

[0003] Current mold management primarily relies on manual recording and simple spreadsheets, which presents the following problems: Incomplete data recording: Data from the entire lifecycle of a mold—from design, manufacturing, use to scrapping—is scattered across different systems, lacking unified recording and traceability capabilities. This results in an inability to accurately analyze mold usage and failure causes, leading to low management efficiency. Outdated maintenance methods: Mold maintenance mainly depends on scheduled maintenance or post-failure repairs, lacking predictive maintenance mechanisms. This leads to: frequent unplanned downtime, impacting production efficiency; high maintenance costs and significant resource waste; and sudden mold failures causing a large number of defective products, affecting product quality. Lack of data-driven decision-making: Existing systems cannot effectively utilize mold operation data for analysis, and cannot accurately predict mold failures and remaining service life. Inaccurate quality control: Insufficient analysis of the correlation between finished product quality and mold status prevents the implementation of refined quality control based on mold status. Summary of the Invention

[0004] To improve the existing system, a digital management and predictive maintenance system for the entire lifecycle of molds based on AI and RFID is provided. This system integrates AI and RFID technologies to accurately assess the health status of molds, provide early warnings of faults, and generate targeted maintenance suggestions. It completes full-process traceability and closed-loop optimization of maintenance data, which greatly improves mold management efficiency, reduces maintenance costs, and extends the service life of molds.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A mold lifecycle digital management and predictive maintenance system based on AI and RFID includes: RFID tag module: Configure each mold with a unique RFID tag to record the mold's unique identifier and basic information; Multi-source data acquisition module: includes a high-precision sensor group and an RFID reading device. The high-precision sensor group includes a vibration sensor, a temperature sensor and a pressure sensor. The RFID reading device reads the mold identification and basic information in real time, and collects mold finished product quality data and mold usage time data. Data transmission module: Utilizes 5G / Industrial IoT network as the data transmission carrier to transmit collected mold identification, basic information, operating status data, finished product quality data, and usage duration data; Cloud-based data processing module: Performs full-process data processing on various types of transmitted data, providing feature data for subsequent mold health status analysis and prediction, and is equipped with a distributed database for mold data storage and retrieval; AI Model Training and Deployment Module: The training set is integrated with historical data of the entire life cycle of the mold and extracted key features. Machine learning and deep learning algorithms are used to build a mold health assessment AI model. After optimization, the model is deployed to the cloud and updated once a day. Mold health assessment module: Input real-time feature data into the mold health assessment AI model, and obtain the mold health assessment result through model analysis and calculation; Fault warning and maintenance module: Receives health assessment results in real time and automatically generates warning information. Based on the predicted fault type, mold health status and historical maintenance data, it generates targeted maintenance suggestions. Full lifecycle traceability module: integrates and stores the collected mold full lifecycle data, provides multi-condition query function, and provides a visual display and traceable query of the entire process data from mold design to scrapping; Maintenance record update module: Enter the actual maintenance information of the mold and synchronize it to the distributed database and the mold's full life cycle data archive. The updated maintenance record data will be fed back to the AI ​​model as a new dataset for subsequent training and optimization.

[0006] Preferably, the RFID tag module specifically includes: Mold Information Writing Unit: During the mold initialization phase, the unique ID, model, specifications, manufacturing date, and basic design standard information of the mold are written in one go. The information corresponds one-to-one with the mold and is permanently associated. Unique Identification Management Unit: Assigns a unique RFID tag code to each mold, establishes a mapping file between the tag code and mold information, and provides a core index for full lifecycle data traceability; Label environment adaptation unit: Based on the mold application scenario and working environment, the label is encapsulated and protected and the installation position is adapted.

[0007] Preferably, the multi-source data acquisition module specifically includes: RFID data reading unit: Reads the unique ID and basic model information from the RFID tag of the mold, verifies the validity of the tag, and binds the mold identity with the collected data; Working condition sensing and acquisition unit: integrates a vibration sensor with a precision of 0.1mg, a temperature sensor with a precision of ±0.5℃, and a pressure sensor to collect vibration frequency, temperature change, and working pressure data in real time during mold operation. The acquisition frequency is synchronized with the mold operation status. Production quality data acquisition unit: Collects data on the dimensional accuracy, surface quality, and defect rate of finished products produced by the mold, and records data on the single run time of the mold and the cumulative number of produced parts; Data calibration unit: timestamps the multi-source data collected by each unit, unify the data collection time base, eliminate invalid collected data, and integrate them into a structured data collection package.

[0008] Preferably, the data transmission module specifically includes: Data encapsulation unit: Standardizes and encapsulates structured acquisition data packets and integrates them into transmission data packets of a unified format; Network transmission unit: Equipped with dual network transmission channels of 5G / Industrial IoT, it automatically switches transmission links according to the network environment of the industrial site and pushes data synchronously to the cloud server; Data verification unit: Performs integrity and consistency verification before and after data transmission. Verifies whether the data packet is lost or tampered with by comparing the check code. Immediately marks the data packet that fails the verification and initiates a retransmission request. Breakpoint resume unit: When the network is interrupted, the data packets to be transmitted are buffered locally, and the data transmission continues from the interruption point after the network is restored.

[0009] Preferably, the cloud data processing module specifically includes: Data receiving unit: receives standardized transmission data packets in real time, classifies them according to the unique ID of the mold, and stores them in a distributed database, supporting the writing and retrieval of massive amounts of data; Data cleaning unit: performs multi-dimensional cleaning on the received raw data, identifies and removes outliers, missing values ​​and duplicate values, completes and corrects incomplete data with deviations within a reasonable range, and verifies the consistency of data format and logic, and filters invalid data; Multi-source data integration unit: Using mold ID and timestamp as dual indexes, it associates and integrates the mold's RFID basic information, working condition sensing data, production quality data, and usage duration data to form a full-dimensional, time-series structured data set for a single mold. Feature extraction and mining unit: Extracts key features related to mold health from the integrated data and performs standardization and normalization processing on the feature data.

[0010] Preferably, the AI ​​model training and deployment module specifically includes: Training dataset construction unit: Integrates feature data and historical data of the entire life cycle of the mold, divides it into training set, validation set and test set, and performs normalization and annotation processing on the data; Algorithm Model Building Unit: Based on the needs of mold failure prediction, an LSTM+CNN hybrid algorithm model is built. The spatial features of the mold operation data are extracted through the CNN architecture, and the time series features of the data are captured through the LSTM architecture to build an AI model for mold health assessment. Model training and optimization unit: The processed dataset is input into the AI ​​model for iterative training. The model accuracy is verified through cross-validation. Parameters are tuned to address model biases. Invalid features are removed to optimize the model structure until the model's fault prediction accuracy is ≥95%. Model Deployment Iteration Unit: Deploy the optimized AI model to the cloud server and iterate and update the model once a day.

[0011] Preferably, the mold health assessment module specifically includes: Model inference and computation unit: calls the mold health assessment AI model deployed in the cloud, inputs real-time feature data into the model for inference analysis, and obtains the current health status result of the mold based on the fault identification and life prediction logic trained by the model. Health indicator generation unit: Based on the model output results, it generates multi-dimensional mold health assessment indicators, quantifies the output of health index and remaining service life from 0 to 100 points, and determines the fault risk level and specific fault type.

[0012] Preferably, the fault early warning and maintenance module specifically includes: Assessment Result Monitoring Unit: Receives real-time output data from the mold health assessment module and continuously monitors the mold health index, remaining service life, failure risk level, and predicted failure type; Threshold judgment and early warning unit: The preset health index, remaining service life safety threshold and fault risk level judgment standard compare the real-time monitoring data with the threshold. If the indicator exceeds the threshold or the risk level reaches medium or high risk, the system early warning mechanism is triggered. Maintenance suggestion unit: Based on the predicted fault types of the early warning model, health status data and historical maintenance cases stored in the system, targeted maintenance measures are matched for different fault types.

[0013] Preferably, the full lifecycle traceability module specifically includes: Data collection unit: Integrates all dimensions of mold lifecycle data according to the mold's unique ID, including design parameters, manufacturing process, RFID basic information, operating conditions, quality data, maintenance records, early warning information, and scrap files; Data index building unit: With mold ID as the core index, an auxiliary index system is built based on time, mold type, fault type, and maintenance frequency. It supports single-condition precise query and multi-condition combined filtering, and also supports fuzzy search. Traceability Results Display Unit: Visualizes and hierarchically displays the retrieved traceability data, including data trend curves, status change history, maintenance ledgers, and early warning records.

[0014] Preferably, the maintenance record update module specifically includes: Maintenance information entry unit: Enters full information on mold maintenance and trial operation test data of the mold after maintenance, adapting to different scenarios of single mold maintenance and batch maintenance of multiple molds; Maintenance data verification unit: performs logical and format verification on the entered maintenance information, checks the matching of mold ID, maintenance content and predicted fault type, verifies the rationality of test data, removes duplicate and incorrect entered information, and reminds users to fill in missing information. Full database data synchronization unit: Updates verified maintenance records to the distributed database and mold lifecycle data archive according to the mold's unique ID; Model Data Feedback Unit: Pushes updated maintenance records to the AI ​​Model Training and Deployment Module as new datasets for iterative model training.

[0015] Compared with the prior art, the advantages of the present invention are: By deeply integrating AI and RFID technologies, this system achieves an integrated upgrade of digital management and predictive maintenance throughout the entire lifecycle of molds, offering significant core advantages while combining practicality and innovation. RFID tags assign a unique identifier to each mold, and combined with multi-source high-precision sensing and reading devices, it enables accurate collection of comprehensive data across all dimensions, including basic mold information, operating conditions, and quality data. Coupled with 5G / Industrial IoT transmission channels, it ensures real-time, complete, and efficient data transmission. Cloud-based data processing and AI model linkage, through a dynamically iterative health assessment model, accurately determine the mold's health status, trigger early fault warnings, and generate targeted maintenance suggestions, effectively avoiding production stoppages caused by sudden failures. Simultaneously, the system enables full-process traceability and data visualization management of molds from design, use, maintenance to scrapping. Closed-loop updates of maintenance records feed back into AI model optimization, significantly improving mold management efficiency, reducing blind maintenance costs, and extending mold lifespan, providing an intelligent, refined, and end-to-end solution for mold management. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the system proposed in this invention; Figure 2 This is a diagram of the RFID tag module proposed in this invention; Figure 3 This is a diagram of the multi-source data acquisition module proposed in this invention; Figure 4 This is a diagram of the data transmission module proposed in this invention; Figure 5 This is a diagram of the cloud data processing module proposed in this invention; Figure 6 This is a diagram of the AI ​​model training and deployment module proposed in this invention; Figure 7 This is a diagram of the mold health assessment module proposed in this invention; Figure 8 This is a diagram of the fault early warning and maintenance module proposed in this invention; Figure 9 This is a diagram of the full lifecycle traceability module proposed in this invention; Figure 10 This is a diagram of the maintenance record update module proposed in this invention. Detailed Implementation

[0017] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0018] See Figure 1 As shown, the mold full lifecycle digital management and predictive maintenance system based on AI and RFID includes: RFID tag module: Configure each mold with a unique RFID tag to record the mold's unique identifier and basic information; Multi-source data acquisition module: includes a high-precision sensor group and an RFID reading device. The high-precision sensor group includes a vibration sensor, a temperature sensor and a pressure sensor. The RFID reading device reads the mold identification and basic information in real time, and collects mold finished product quality data and mold usage time data. Data transmission module: Utilizes 5G / Industrial IoT network as the data transmission carrier to transmit collected mold identification, basic information, operating status data, finished product quality data, and usage duration data; Cloud-based data processing module: Performs full-process data processing on various types of transmitted data, providing feature data for subsequent mold health status analysis and prediction, and is equipped with a distributed database for mold data storage and retrieval; AI Model Training and Deployment Module: The training set is integrated with historical data of the entire life cycle of the mold and extracted key features. Machine learning and deep learning algorithms are used to build a mold health assessment AI model. After optimization, the model is deployed to the cloud and updated once a day. Mold health assessment module: Input real-time feature data into the mold health assessment AI model, and obtain the mold health assessment result through model analysis and calculation; Fault warning and maintenance module: Receives health assessment results in real time and automatically generates warning information. Based on the predicted fault type, mold health status and historical maintenance data, it generates targeted maintenance suggestions. Full lifecycle traceability module: integrates and stores the collected mold full lifecycle data, provides multi-condition query function, and provides a visual display and traceable query of the entire process data from mold design to scrapping; Maintenance record update module: Enter the actual maintenance information of the mold and synchronize it to the distributed database and the mold's full life cycle data archive. The updated maintenance record data will be fed back to the AI ​​model as a new dataset for subsequent training and optimization.

[0019] See Figure 2 As shown, the RFID tag module specifically includes: Mold Information Writing Unit: During the mold initialization phase, the unique ID, model, specifications, manufacturing date, and basic design standard information of the mold are written in one go. The information corresponds one-to-one with the mold and is permanently associated. Unique Identification Management Unit: Assigns a unique RFID tag code to each mold, establishes a mapping file between the tag code and mold information, and provides a core index for full lifecycle data traceability; Label environment adaptation unit: Based on the mold application scenario and working environment, the label is encapsulated and protected and the installation position is adapted.

[0020] Specifically, the tags are configured in a one-to-one correspondence manner, with each mold individually equipped with an RFID tag. Installation and basic information writing are completed during the mold initialization phase. Tag data storage is divided into two parts: initial permanent writing and incremental writing throughout the entire lifecycle. During the initialization phase, core basic information such as the mold's unique ID, mold model, specifications, manufacturer, production date, and design lifespan are written. This type of information is stored in an encrypted mode and cannot be tampered with, establishing a unique digital identity file for the mold. Throughout the entire lifecycle, dynamic basic information such as the mold's first use time, installation position, and initial maintenance node can be incrementally written through RFID reading devices. All stored information is bound to the mold's unique ID, providing an identity matching basis for subsequent multi-source data association and integration.

[0021] See Figure 3 As shown, the multi-source data acquisition module specifically includes: RFID data reading unit: Reads the unique ID and basic model information from the RFID tag of the mold, verifies the validity of the tag, and binds the mold identity with the collected data; Working condition sensing and acquisition unit: integrates a vibration sensor with a precision of 0.1mg, a temperature sensor with a precision of ±0.5℃, and a pressure sensor to collect vibration frequency, temperature change, and working pressure data in real time during mold operation. The acquisition frequency is synchronized with the mold operation status. Production quality data acquisition unit: Collects data on the dimensional accuracy, surface quality, and defect rate of finished products produced by the mold, and records data on the single run time of the mold and the cumulative number of produced parts; Data calibration unit: timestamps the multi-source data collected by each unit, unify the data collection time base, eliminate invalid collected data, and integrate them into a structured data collection package.

[0022] Specifically, the module simultaneously connects to production line testing and counting equipment, automatically collecting finished product quality data from mold production, including finished product defect rate, dimensional accuracy, appearance pass rate, and actual mold usage time data, including cumulative working hours, cumulative production quantity, and single operation time, eliminating the need for manual data entry and reducing human error. All acquisition devices are connected to an embedded data acquisition terminal, which supports industrial common interfaces such as RS485 and Modbus, enabling unified format and aggregation of multi-source data. It establishes a real-time data archive for each mold based on its unique mold ID, ensuring accurate matching of identification information, operating data, and finished product data in the time dimension.

[0023] See Figure 4 As shown, the data transmission module specifically includes: Data encapsulation unit: Standardizes and encapsulates structured acquisition data packets and integrates them into transmission data packets of a unified format; Network transmission unit: Equipped with dual network transmission channels of 5G / Industrial IoT, it automatically switches transmission links according to the network environment of the industrial site and pushes data synchronously to the cloud server; Data verification unit: Performs integrity and consistency verification before and after data transmission. Verifies whether the data packet is lost or tampered with by comparing the check code. Immediately marks the data packet that fails the verification and initiates a retransmission request. Breakpoint resume unit: When the network is interrupted, the data packets to be transmitted are buffered locally, and the data transmission continues from the interruption point after the network is restored.

[0024] Specifically, at the communication level, a dual-module redundancy design of 5G + Industrial IoT is adopted. When the 5G signal is interrupted or the network is weak at the production site, the system will automatically and seamlessly switch to the backup module to ensure uninterrupted data transmission. Data transmission adopts the TLS / SSL industrial-grade end-to-end encryption protocol to encrypt all transmitted data, preventing data from being tampered with or stolen during transmission, and ensuring the security of core information such as mold operation data and quality data. The module has a built-in comprehensive transmission control mechanism with breakpoint resume function. After the network is restored, it will automatically resume transmission from the breakpoint to avoid data loss. At the same time, it supports data priority transmission, marking high-priority data such as abnormal mold health index and high fault risk and transmitting them first, while ordinary operation data is transmitted at the normal rate to ensure the timeliness of fault warning data.

[0025] See Figure 5 As shown, the cloud data processing module specifically includes: Data receiving unit: receives standardized transmission data packets in real time, classifies them according to the unique ID of the mold, and stores them in a distributed database, supporting the writing and retrieval of massive amounts of data; Data cleaning unit: performs multi-dimensional cleaning on the received raw data, identifies and removes outliers, missing values ​​and duplicate values, completes and corrects incomplete data with deviations within a reasonable range, and verifies the consistency of data format and logic, and filters invalid data; Multi-source data integration unit: Using mold ID and timestamp as dual indexes, it associates and integrates the mold's RFID basic information, working condition sensing data, production quality data, and usage duration data to form a full-dimensional, time-series structured data set for a single mold. Feature extraction and mining unit: Extracts key features related to mold health from the integrated data and performs standardization and normalization processing on the feature data.

[0026] Specifically, the data preprocessing submodule is the core processing step, which is divided into two refined steps: data cleaning and data integration. The data cleaning step is based on the 3σ principle of the mold industry scenario, which identifies and removes abrupt outliers in the operating data such as vibration and temperature. It uses the interpolation method of the same mold's historical time series data to complete the missing values ​​caused by temporary sensor failures, and deletes duplicate data collected by timestamp and mold ID. The data integration step uses the mold's unique ID as the core index and the timestamp as the correlation dimension to fuse RFID identification basic data, sensor operating status data, finished product quality data, and mold usage time data from multiple sources to form a full-dimensional integrated time series data archive for a single mold. At the same time, it standardizes the heterogeneous equipment data into a structured format of "mold ID + timestamp + data index + value". The feature extraction submodule combines mold failure mechanisms with practical industry operation and maintenance experience to perform targeted key feature mining on preprocessed structured data. For operational data such as vibration, temperature, and pressure, it extracts core features such as vibration frequency distribution, temperature change rate, pressure fluctuation amplitude, and duration of vibration peak. For finished product quality data, it extracts features such as finished product defect rate, correlation between defect type and mold operating parameters, and number of defective parts per unit time. For basic mold operation data, it extracts features such as cumulative usage time, continuous duration of a single operation, and maintenance interval. At the same time, all extracted features are normalized to eliminate the dimensional differences between different indicators.

[0027] See Figure 6 As shown, the AI ​​model training and deployment module specifically includes: Training dataset construction unit: Integrates feature data and historical data of the entire life cycle of the mold, divides it into training set, validation set and test set, and performs normalization and annotation processing on the data; Algorithm Model Building Unit: Based on the needs of mold failure prediction, an LSTM+CNN hybrid algorithm model is built. The spatial features of the mold operation data are extracted through the CNN architecture, and the time series features of the data are captured through the LSTM architecture to build an AI model for mold health assessment. Model training and optimization unit: The processed dataset is input into the AI ​​model for iterative training. The model accuracy is verified through cross-validation. Parameters are tuned to address model biases. Invalid features are removed to optimize the model structure until the model's fault prediction accuracy is ≥95%. Model Deployment Iteration Unit: Deploy the optimized AI model to the cloud server and iterate and update the model once a day.

[0028] Specifically, a two-layer architecture of "CNN-LSTM hybrid deep learning layer + industrial rule calibration layer" is constructed. The CNN layer contains 3 convolutional layers, 2 max pooling layers and ReLU activation layers, which are responsible for extracting spatial local features of data such as mold vibration and temperature; the LSTM layer has 2 hidden layers, which receive the spatial features output by the CNN layer and capture the temporal series features of the data; the industrial rule calibration layer, as the top-level module, is directly connected to the output of the LSTM layer, embedding mold operation and maintenance expert experience rules to calibrate the original output of the model. The training process is divided into four stages: First, data preprocessing is performed, dividing the cloud-processed feature data into training, validation, and test sets in a 7:2:1 ratio. The SMOTE algorithm is used to enhance low-frequency fault data and supplement features from similar operating conditions. Second, core parameters are initialized. First, the CNN layer is pre-trained separately to lock in spatial feature extraction capabilities, then the LSTM layer is trained to capture temporal features, and finally, the industrial rule layer is fused for joint training. Third, optimization is achieved through 5-fold cross-validation, testing the combination of learning rate and batch size parameters. The optimal parameters are selected based on fault prediction accuracy, false positive rate, and false negative rate as core indicators. Fourth, the model is lightweighted by pruning redundant neurons and compressing the model size to less than 500MB to improve inference efficiency. For stamping dies, the weights of the CNN layer for extracting pressure fluctuations and stress distribution features are increased to enhance the ability to identify wear and chipping faults. For injection molds, the sensitivity of the LSTM layer for capturing melt temperature and mold cavity pressure time-series features is improved to focus on identifying mold sticking and cooling system faults. At the same time, historical fault data of different molds are mapped to model classification labels to make the model output fit the actual operation and maintenance scenario. The model's input and output settings form a strong correlation logic: the input dimension includes three types of structured data: real-time operating characteristics of the mold, basic mold attributes, and historical fault and maintenance records; the output dimension is a standardized evaluation result, including a health index of 0-100 points, remaining service life, low / medium / high fault risk levels, and specific fault types. The input features and output results are correlated through weight allocation to ensure that the output results are traceable and interpretable.

[0029] See Figure 7 As shown, the mold health assessment module specifically includes: Model inference and computation unit: calls the mold health assessment AI model deployed in the cloud, inputs real-time feature data into the model for inference analysis, and obtains the current health status result of the mold based on the fault identification and life prediction logic trained by the model. Health indicator generation unit: Based on the model output results, it generates multi-dimensional mold health assessment indicators, quantifies the output of health index and remaining service life from 0 to 100 points, and determines the fault risk level and specific fault type.

[0030] Specifically, the intelligent health status computing unit undertakes the core assessment and calculation work, and completely inputs the fused input dataset into the called mold health assessment AI model, which then completes the comprehensive reasoning and quantitative calculation of multi-dimensional features. The standardized assessment result generation unit transforms the raw results of model inference calculations into standardized and visualized assessment results that can be directly applied in industrial scenarios. It strictly outputs a health index of 0-100 points across four core dimensions, with higher scores indicating better mold condition; remaining service life measured in working hours or production units; low, medium, and high failure risk levels categorized as <20%, 20%-50%, and >50%; and predictions of specific failure types such as mold wear, component loosening, and damage to vulnerable parts. All assessment results are permanently bound to the mold's unique ID with a calculation timestamp and model version identifier. Formula for predicting remaining useful life: ; in, For the remaining service life, To design rated life, For mold health index, The aging index, This represents the total usage time. The results caching and multi-terminal synchronization unit will cache the standardized evaluation results generated in real time to the cloud high-speed cache database, establish a time-series archive of evaluation results according to the unique ID of the mold, and support subsequent health status trend analysis and quick retrieval. At the same time, the evaluation results will be synchronously pushed to the fault warning and maintenance suggestion module and the system visual operation interface to realize real-time sharing of evaluation results and provide data support for fault warning judgment and on-site visual monitoring.

[0031] See Figure 8 As shown, the fault early warning and maintenance module specifically includes: Assessment Result Monitoring Unit: Receives real-time output data from the mold health assessment module and continuously monitors the mold health index, remaining service life, failure risk level, and predicted failure type; Threshold judgment and early warning unit: The preset health index, remaining service life safety threshold and fault risk level judgment standard compare the real-time monitoring data with the threshold. If the indicator exceeds the threshold or the risk level reaches medium or high risk, the system early warning mechanism is triggered. Maintenance suggestion unit: Based on the predicted fault types of the early warning model, health status data and historical maintenance cases stored in the system, targeted maintenance measures are matched for different fault types.

[0032] Specifically, the system receives standardized assessment results from the mold health assessment module in real time, establishes an independent real-time monitoring queue based on the mold's unique ID, and continuously monitors the health index, remaining service life, fault risk level, and fault type prediction results of each mold at the millisecond level and performs multi-condition superposition judgment. If any indicator reaches the preset threshold, the early warning judgment mechanism is triggered. At the same time, the system embeds trend prediction logic to issue early warnings for molds with continuously declining health index and gradually increasing fault risk level, thus achieving dual monitoring of "threshold-triggered early warning + trend prediction early warning". Based on the early warning judgment results, standardized and structured early warning information is generated by associating it with the mold's unique ID. The information includes core elements: basic mold identification information, current health assessment full data, specific threshold type for early warning triggering, early warning classification results, predicted fault type and risk development trend. High-risk early warning information is specially highlighted in red and marked with emphasis. All early warning information is given a unique early warning code and trigger timestamp, and is permanently bound to the mold's full life cycle data. Based on early warning information and the system's full data, practical maintenance suggestions are generated. The core of the suggestions combines four dimensions of data: mold predicted failure type, historical maintenance records and operation and maintenance patterns of the mold, industry best maintenance cases of similar molds, and production line production plan and workstation allocation. Targeted suggestions are generated for different failure types, while also specifying maintenance priority, recommended maintenance duration, required core parts and key operation points.

[0033] See Figure 9 As shown, the full lifecycle traceability module specifically includes: Data collection unit: Integrates all dimensions of mold lifecycle data according to the mold's unique ID, including design parameters, manufacturing process, RFID basic information, operating conditions, quality data, maintenance records, early warning information, and scrap files; Data index building unit: With mold ID as the core index, an auxiliary index system is built based on time, mold type, fault type, and maintenance frequency. It supports single-condition precise query and multi-condition combined filtering, and also supports fuzzy search. Traceability Results Display Unit: Visualizes and hierarchically displays the retrieved traceability data, including data trend curves, status change history, maintenance ledgers, and early warning records.

[0034] Specifically, using the mold's unique ID as a global index, the system connects to all system units in real time, comprehensively collecting data from all stages of the mold's lifecycle. This includes real-time operational data, production volume, and finished product quality data during the usage stage; health assessment records, fault warning information, maintenance operation details, and replacement component information during the maintenance and early warning stage; and data such as scrapping reasons, inspection reports, and residual value assessments during the scrapping stage. This unit incorporates a data association and verification mechanism, precisely binding data from each module to the mold ID based on timestamps, eliminating duplicate and invalid data, and achieving second-level synchronization of updated data across modules, ensuring the integrity, timeliness, and consistency of traceability data.

[0035] See Figure 10 As shown, the maintenance record update module specifically includes: Maintenance information entry unit: Enters full information on mold maintenance and trial operation test data of the mold after maintenance, adapting to different scenarios of single mold maintenance and batch maintenance of multiple molds; Maintenance data verification unit: performs logical and format verification on the entered maintenance information, checks the matching of mold ID, maintenance content and predicted fault type, verifies the rationality of test data, removes duplicate and incorrect entered information, and reminds users to fill in missing information. Full database data synchronization unit: Updates verified maintenance records to the distributed database and mold lifecycle data archive according to the mold's unique ID; Model Data Feedback Unit: Pushes updated maintenance records to the AI ​​Model Training and Deployment Module as new datasets for iterative model training.

[0036] Specifically, the standardized processing of maintenance data incorporates a dual verification mechanism: field verification automatically identifies issues such as missing required fields and incorrect data formats, providing real-time pop-up alerts to ensure the completeness of entered information; logical verification correlates and matches the entered actual fault type and handling method with the warning information from the fault warning and maintenance suggestion module, verifying the correlation between maintenance behavior and warning prompts, while also checking the compatibility of replacement parts with mold models to avoid invalid data entry. After successful verification, this unit transforms unstructured entered information and attachment data into standardized structured data, creating an index based on "unique mold ID + maintenance record code," and adding an entry timestamp, operator identifier, and data verification code. According to the system's preset daily incremental update frequency, the structured maintenance record data is automatically synchronized to the AI ​​model training and deployment module as a new dataset for model iterative training. It focuses on integrating core data such as the matching degree between actual fault types and predicted fault types, the effectiveness of maintenance handling methods, and the operation and maintenance cycle of different faults, providing practical basis for model parameter optimization and fault prediction accuracy improvement.

[0037] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0038] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0039] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A mold lifecycle digital management and predictive maintenance system based on AI and RFID, characterized in that, include: RFID tag module: Configure each mold with a unique RFID tag to record the mold's unique identifier and basic information; Multi-source data acquisition module: includes a high-precision sensor group and an RFID reading device. The high-precision sensor group includes a vibration sensor, a temperature sensor and a pressure sensor. The RFID reading device reads the mold identification and basic information in real time, and collects mold finished product quality data and mold usage time data. Data transmission module: Utilizes 5G / Industrial IoT network as the data transmission carrier to transmit collected mold identification, basic information, operating status data, finished product quality data, and usage duration data; Cloud-based data processing module: Performs full-process data processing on various types of transmitted data, providing feature data for subsequent mold health status analysis and prediction, and is equipped with a distributed database for mold data storage and retrieval; AI Model Training and Deployment Module: The training set is integrated with historical data of the entire life cycle of the mold and extracted key features. Machine learning and deep learning algorithms are used to build a mold health assessment AI model. After optimization, the model is deployed to the cloud and updated once a day. Mold health assessment module: Input real-time feature data into the mold health assessment AI model, and obtain the mold health assessment result through model analysis and calculation; Fault warning and maintenance module: Receives health assessment results in real time and automatically generates warning information. Based on the predicted fault type, mold health status and historical maintenance data, it generates targeted maintenance suggestions. Full lifecycle traceability module: integrates and stores the collected mold full lifecycle data, provides multi-condition query function, and provides a visual display and traceable query of the entire process data from mold design to scrapping; Maintenance record update module: Enter the actual maintenance information of the mold and synchronize it to the distributed database and the mold's full life cycle data archive. The updated maintenance record data will be fed back to the AI ​​model as a new dataset for subsequent training and optimization.

2. The AI ​​and RFID-based mold lifecycle digital management and predictive maintenance system according to claim 1, characterized in that, The RFID tag module specifically includes: Mold Information Writing Unit: During the mold initialization phase, the unique ID, model, specifications, manufacturing date, and basic design standard information of the mold are written in one go. The information corresponds one-to-one with the mold and is permanently associated. Unique Identification Management Unit: Assigns a unique RFID tag code to each mold, establishes a mapping file between the tag code and mold information, and provides a core index for full lifecycle data traceability; Label environment adaptation unit: Based on the mold application scenario and working environment, the label is encapsulated and protected and the installation position is adapted.

3. The AI ​​and RFID-based mold lifecycle digital management and predictive maintenance system according to claim 1, characterized in that, The multi-source data acquisition module specifically includes: RFID data reading unit: Reads the unique ID and basic model information from the RFID tag of the mold, verifies the validity of the tag, and binds the mold identity with the collected data; Working condition sensing and acquisition unit: integrates a vibration sensor with a precision of 0.1mg, a temperature sensor with a precision of ±0.5℃, and a pressure sensor to collect vibration frequency, temperature change, and working pressure data in real time during mold operation. The acquisition frequency is synchronized with the mold operation status. Production quality data acquisition unit: Collects data on the dimensional accuracy, surface quality, and defect rate of finished products produced by the mold, and records data on the single run time of the mold and the cumulative number of produced parts; Data calibration unit: timestamps the multi-source data collected by each unit, unify the data collection time base, eliminate invalid collected data, and integrate them into a structured data collection package.

4. The AI ​​and RFID-based mold lifecycle digital management and predictive maintenance system according to claim 1, characterized in that, The data transmission module specifically includes: Data encapsulation unit: Standardizes and encapsulates structured acquisition data packets and integrates them into transmission data packets of a unified format; Network transmission unit: Equipped with dual network transmission channels of 5G / Industrial IoT, it automatically switches transmission links according to the network environment of the industrial site and pushes data synchronously to the cloud server; Data verification unit: Performs integrity and consistency verification before and after data transmission. Verifies whether the data packet is lost or tampered with by comparing the check code. Immediately marks the data packet that fails the verification and initiates a retransmission request. Breakpoint resume unit: When the network is interrupted, the data packets to be transmitted are buffered locally, and the data transmission continues from the interruption point after the network is restored.

5. The AI ​​and RFID-based mold lifecycle digital management and predictive maintenance system according to claim 1, characterized in that, The cloud data processing module specifically includes: Data receiving unit: receives standardized transmission data packets in real time, classifies them according to the unique ID of the mold, and stores them in a distributed database, supporting the writing and retrieval of massive amounts of data; Data cleaning unit: performs multi-dimensional cleaning on the received raw data, identifies and removes outliers, missing values ​​and duplicate values, completes and corrects incomplete data with deviations within a reasonable range, and verifies the consistency of data format and logic, and filters invalid data; Multi-source data integration unit: Using mold ID and timestamp as dual indexes, it associates and integrates the mold's RFID basic information, working condition sensing data, production quality data, and usage duration data to form a full-dimensional, time-series structured data set for a single mold. Feature extraction and mining unit: Extracts key features related to mold health from the integrated data and performs standardization and normalization processing on the feature data.

6. The AI ​​and RFID-based mold lifecycle digital management and predictive maintenance system according to claim 1, characterized in that, The AI ​​model training and deployment module specifically includes: Training dataset construction unit: Integrates feature data and historical data of the entire life cycle of the mold, divides it into training set, validation set and test set, and performs normalization and annotation processing on the data; Algorithm Model Building Unit: Based on the needs of mold failure prediction, an LSTM+CNN hybrid algorithm model is built. The spatial features of the mold operation data are extracted through the CNN architecture, and the time series features of the data are captured through the LSTM architecture to build an AI model for mold health assessment. Model training and optimization unit: The processed dataset is input into the AI ​​model for iterative training. The model accuracy is verified through cross-validation. Parameters are tuned to address model biases. Invalid features are removed to optimize the model structure until the model's fault prediction accuracy is ≥95%. Model Deployment Iteration Unit: Deploy the optimized AI model to the cloud server and iterate and update the model once a day.

7. The AI ​​and RFID-based mold lifecycle digital management and predictive maintenance system according to claim 1, characterized in that, The mold health assessment module specifically includes: Model inference and computation unit: calls the mold health assessment AI model deployed in the cloud, inputs real-time feature data into the model for inference analysis, and obtains the current health status result of the mold based on the fault identification and life prediction logic trained by the model. Health indicator generation unit: Based on the model output results, it generates multi-dimensional mold health assessment indicators, quantifies the output of health index and remaining service life from 0 to 100 points, and determines the fault risk level and specific fault type.

8. The AI ​​and RFID-based mold lifecycle digital management and predictive maintenance system according to claim 1, characterized in that, The fault early warning and maintenance module specifically includes: Assessment Result Monitoring Unit: Receives real-time output data from the mold health assessment module and continuously monitors the mold health index, remaining service life, failure risk level, and predicted failure type; Threshold judgment and early warning unit: The preset health index, remaining service life safety threshold and fault risk level judgment standard compare the real-time monitoring data with the threshold. If the indicator exceeds the threshold or the risk level reaches medium or high risk, the system early warning mechanism is triggered. Maintenance suggestion unit: Based on the predicted fault types of the early warning model, health status data and historical maintenance cases stored in the system, targeted maintenance measures are matched for different fault types.

9. The AI ​​and RFID-based mold lifecycle digital management and predictive maintenance system according to claim 1, characterized in that, The full lifecycle traceability module specifically includes: Data collection unit: Integrates all dimensions of mold lifecycle data according to the mold's unique ID, including design parameters, manufacturing process, RFID basic information, operating conditions, quality data, maintenance records, early warning information, and scrap files; Data index building unit: With mold ID as the core index, an auxiliary index system is built based on time, mold type, fault type, and maintenance frequency. It supports single-condition precise query and multi-condition combined filtering, and also supports fuzzy search. Traceability Results Display Unit: Visualizes and hierarchically displays the retrieved traceability data, including data trend curves, status change history, maintenance ledgers, and early warning records.

10. The AI ​​and RFID-based mold lifecycle digital management and predictive maintenance system according to claim 1, characterized in that, The maintenance record update module specifically includes: Maintenance information entry unit: Enters full information on mold maintenance and trial operation test data of the mold after maintenance, adapting to different scenarios of single mold maintenance and batch maintenance of multiple molds; Maintenance data verification unit: performs logical and format verification on the entered maintenance information, checks the matching of mold ID, maintenance content and predicted fault type, verifies the rationality of test data, removes duplicate and incorrect entered information, and reminds users to fill in missing information. Full database data synchronization unit: Updates verified maintenance records to the distributed database and mold lifecycle data archive according to the mold's unique ID; Model Data Feedback Unit: Pushes updated maintenance records to the AI ​​Model Training and Deployment Module as new datasets for iterative model training.

Images