A real-time data driven artificial intelligence smart TPM predictive maintenance system

Abstract

A real-time data-driven artificial intelligence SmartTPM predictive maintenance system, comprising a data source layer, a WebAPI service layer, a database entity layer, a component life assessment layer, an AI large model decision layer, an application display layer, a periodic rolling update layer, and an execution feedback closed loop, wherein a component life digital twin object is constructed, multi-source operation and maintenance data is fused to quantify health index, residual life and risk level, combined with AI large model output standardized maintenance strategy, spare parts and cycle optimization scheme, adaptive rolling update maintenance cycle, through device, component, maintenance work order, fault repair, spare parts inventory, production line visual board full business link, realizing the transformation from passive after-maintenance to predictive intelligent maintenance closed loop.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing industrial internet equipment operation and maintenance technology, and in particular to a real-time data-driven artificial intelligence SmartTPM predictive maintenance system. Background Technology

[0002] In intelligent manufacturing production line operations, the operational stability of production equipment directly determines the production cycle time, product yield, order delivery cycle, and spare parts maintenance costs. Currently, manufacturing enterprises generally implement digital maintenance management systems to digitize equipment ledgers, daily inspections, regular maintenance, fault reporting work orders, and spare parts warehousing processes, replacing traditional paper ledgers and reducing offline manual coordination costs. However, existing traditional maintenance systems have inherent defects in their underlying architecture and operational logic.

[0003] Traditional maintenance systems rely on equipment-level management, fixed calendar cycle triggering, and manual experience-based repair. The system presets fixed maintenance cycles such as 7, 30, and 90 days, automatically generating maintenance tasks. Repair personnel depend on their personal experience to determine the need for cleaning, inspection, and replacement of components. Managers can only check whether maintenance tasks are due and whether work orders are closed through a visual dashboard. This model only digitizes the maintenance process and cannot differentiate the lifespan differences of various worn components within the same equipment.

[0004] Core consumable components in production line equipment, such as nozzles, cylinders, sensors, motors, and guide rails, are affected by multiple factors, including equipment load, operating time, workshop temperature and humidity, maintenance execution quality, failure frequency, self-inspection anomalies, and spare parts replacement behavior, resulting in varying wear rates. A single, fixed time cycle has significant limitations; high-frequency consumable components are prone to under-maintenance, leading to sudden downtime and product defects. Conversely, mandatory periodic maintenance of low-consumption components results in over-maintenance, wasting labor and spare parts, and significantly increasing operating costs.

[0005] Existing similar digital maintenance systems only have basic ledger and work order processing capabilities, and have the following core shortcomings: 1. The management granularity is too coarse, with the whole equipment and maintenance work orders as the core of management, without being refined to the component level, making it impossible to accurately locate high-loss and high-risk parts; 2. Maintenance triggers rely on fixed calendar cycles, making it impossible to assess the true remaining service life by combining the real-time operating status of the equipment and the degree of component wear; 3. There is no adaptive cycle correction mechanism. After early maintenance, delayed maintenance, fault repair, or spare parts replacement, the mechanical maintenance date is only postponed, and the maintenance cycle cannot be dynamically corrected. 4. The repair plan relies entirely on the personal experience of senior repair personnel. The root cause of the fault, standardized repair procedures, spare parts list, and cycle optimization strategy cannot be automatically accumulated and reused. 5. The visual dashboard has a single information dimension, only showing the maintenance due date and the number of faults, and cannot intuitively show the health status of components, remaining usable life, fault risk causes and corresponding handling solutions; 6. Spare parts inventory is disconnected from equipment component lifespan risk. Inventory only records the material inflow and outflow, and cannot predict spare parts shortage risk in advance based on component lifespan. 7. There is no unified closed-loop operation and maintenance link. Component life assessment, maintenance decision-making, maintenance execution, data feedback, and knowledge accumulation are independent of each other, making it impossible to form a continuously self-optimizing intelligent maintenance system. Summary of the Invention

[0006] To address the problems and technical requirements of existing technologies, the purpose of this invention is to provide a real-time data-driven AI-based SmartTPM predictive maintenance system. This system constructs digital twin objects of component lifespan, integrates multi-source maintenance data to quantitatively calculate health indices, remaining lifespan, and risk levels, and combines AI large-scale models to output standardized maintenance strategies, spare parts, and cycle optimization schemes. It adaptively and continuously updates maintenance cycles, and through the entire business chain of equipment, components, maintenance work orders, fault reporting, spare parts inventory, and production line visualization dashboards, it achieves a transformation from passive, reactive maintenance to a predictive intelligent maintenance closed-loop system.

[0007] To achieve the above objectives, the present invention employs the following technical solution: A real-time data-driven AI-powered SmartTPM predictive maintenance system includes a data source layer, a WebAPI service layer, a database entity layer, a component life assessment layer, an AI large model decision-making layer, an application display layer, a periodic rolling update layer, and an execution feedback closed loop. A method for implementing a real-time data-driven, AI-based SmartTPM predictive maintenance system includes the following steps: Step S1: The data source layer provides basic data to the WebAPI service layer through business calls. The basic data includes equipment templates, modules, components, maintenance items, equipment instances, production line locations, self-inspection records, fault work orders, spare parts ledgers, and inventory flow. Step S2: The WebAPI service layer receives business calls through the equipment management service, production line dashboard service, fault reporting service, and spare parts inventory service, and provides business data to the component life assessment layer through aggregation calculation; Step S3: The database entity layer stores data and provides data support for the component life assessment layer. The stored data includes maintenance plan, maintenance execution record, self-inspection record, fault work order, and spare parts inventory flow. Step S4: The component life assessment layer constructs a life feature vector X based on multi-source data, calculates the life decay index D, remaining life RUL, health index HI and risk level, and outputs the status to the application display layer, and outputs the life status and risk context to the AI ​​big model decision layer. Step S5: The AI ​​big model decision layer generates maintenance steps, spare parts suggestions, risk explanations and cycle adjustment suggestions based on fault semantic understanding, historical maintenance records, component life status, spare parts inventory and knowledge base, and outputs the strategy suggestions to the application display layer and the cycle suggestions to the cycle rolling update layer. Step S6: The cycle rolling update layer dynamically calculates the cycle adjustment factor CycleFactor, the next maintenance due time, and the cycle assignment based on the actual maintenance completion time, component life assessment results, and cycle recommendations, and outputs the update task to the application display layer. Step S7: The application display layer will output the status received from the component life assessment layer to the personal workbench, data overview, equipment management, fault reporting, spare parts inventory and production line dashboard, and output work assignment / display to the execution feedback loop; Step S8: Execute the feedback loop to write back the records to the database entity layer of step S3. The content of the record write-back includes maintenance completion, fault handling, component replacement, and inventory in / out results.

[0008] The data source layer in step S1 includes the following steps: Step S101: Establish the equipment template layer structure. The equipment template layer includes equipment templates, modules, components, and maintenance items. The equipment template describes the standard structure of a certain type of equipment. The module divides the internal functional areas of the equipment. The component records the component name, component baseline life L0, and component importance level of the maintainable object. The maintenance item records the maintenance content, maintenance cycle, and early warning lead time of the corresponding component. Step S102: Configure baseline life parameters for each component. The baseline life parameters include design life days, default maintenance cycle, component importance level, early warning lead time, and optional spare parts replacement strategy. Step S103: Create an equipment instance, bind the equipment instance to the production line location and equipment template, and map the template component to a component life digital twin object under the equipment instance. The component life digital twin object includes the following fields: component identifier, equipment identifier, module name, component name, component baseline life L0, number of days used or equivalent usage time U, last maintenance completion time Tm, next maintenance due time Td, number of failures F, number of self-inspection anomalies Q, overdue days O, spare parts replacement status P, health index HI, remaining life RUL, and risk level RiskLevel. The RiskLevel includes Normal, Attention, Warning, Abnormal, and Frozen status; Step S104: Construct a digital twin model of component lifespan, including an equipment template layer, an equipment instance layer, a maintenance plan snapshot layer, a component lifespan status layer, a component replacement record layer, and a dashboard display layer; The equipment instance layer records the equipment actually put into production, including equipment number, production line, activation time and Kanban coordinates. When an equipment instance is bound to an equipment template, a corresponding maintenance plan snapshot is automatically generated based on the modules, components and maintenance items in the template. The maintenance plan snapshot saves the maintenance status of the device instance within the current cycle. The maintenance status includes device identifier, module name, component name, last completion time, next maintenance due time, cycle maintenance due time, and maintenance item content. The component life status layer is generated by the component life digital twin object with "equipment instance + component + maintenance plan" as the basic unit, and stores the component's baseline life L0, number of days used or equivalent usage time U, the time of the most recent maintenance completion Tm, the time of the next maintenance due Td, the number of failures F, the number of self-inspection anomalies Q, the number of overdue days O, the spare parts replacement status P, the health index HI, the remaining life RUL, and the risk level RiskLevel. The component replacement record layer records the replacement time, replaced component, spare parts information, operator and remarks. When the system identifies that a critical component has been replaced, it resets or partially restores the component's lifespan based on the replacement record and recalculates the next maintenance time Td. The aforementioned dashboard display layer displays a list of component lifespans, remaining days, health index (HI), and status color based on the component lifespan digital twin object.

[0009] In step S103, the component baseline lifespan L0 is initialized. If the component is configured with a component lifespan in days (ComponentLifeDays), then the component baseline lifespan L0 = ComponentLifeDays. If there is a maintenance item cycle (IntervalDays) under the component, then L0 = the shortest critical maintenance cycle or the weighted maintenance cycle. Otherwise, L0 = the system default lifespan parameter. When multiple maintenance items apply to the same component simultaneously, a weighted fusion formula can be used:

[0010] in, , , For weight parameters, This is the shortest cycle among the maintenance items related to the component. The average cycle of maintenance items related to the component is given by CriticalLevelFactor, which is a correction factor obtained from Level 1 maintenance, Level 2 maintenance, Level 3 maintenance, or component importance level.

[0011] The component life assessment layer in step S4 includes the following component life assessment steps: Step S401: Read the equipment instance and template structure, read the hierarchical relationship between equipment instances, equipment templates, modules, components and maintenance items, and determine the equipment and component objects to be assessed for life. Step S402: Summarize the maintenance plan snapshot and read the maintenance plan information corresponding to the component, including the number of days of the maintenance cycle, the next maintenance due date, the warning lead time, the last maintenance completion time and the cycle maintenance due date; Step S403: Read the component lifespan start point. Determine the start time for component lifespan calculation based on the equipment activation time or the most recent component replacement time. If a component replacement record exists, use the replacement time as the new lifespan cycle start point. Step S404: Calculate the lifespan baseline and the number of days used. Determine the component baseline lifespan L0 based on the component lifespan days, maintenance item cycle, or replacement record lifespan snapshot, and calculate the number of days used or equivalent usage time U based on the current date, activation time, maintenance records, and replacement records. Step S405: Construct a lifespan feature vector X based on the number of days used or equivalent usage time U, the current maintenance cycle consumption rate C, the number of overdue days O, the number of failures F, the number of self-inspection anomalies Q, the most recent maintenance quality score M, the spare parts replacement status P, the impact of early warning lead time W, the equipment status or production line importance level S, and the abnormal semantic risk A inferred by the AI ​​big model from the text records. Step S406: Calculate the additional attenuation index D based on cycle consumption, overdue penalty, fault penalty, self-inspection anomaly, maintenance quality, equipment importance and AI semantic risk factors; Step S407: Based on the component's reference life L0 and equivalent service life Life decay index D, life recovery amount due to effective maintenance And the life reset amount caused by spare parts replacement or component replacement Calculate the remaining life expectancy (RUL) and convert it into a health index (HI), while also outputting the risk level. Step S408: Determine the risk level when... And remaining lifespan A value greater than the warning window is considered normal, and the current maintenance cycle should be maintained. or remaining lifespan Enter the warning window as a watchlist. Multiple anomalies may trigger an early warning, providing advance reminders or shortening the maintenance cycle. And remaining lifespan A serious malfunction may be detected as an anomaly, triggering a repair / maintenance request, shutdown for inspection, or spare parts preparation. If the equipment is out of service or not effective, it will be frozen. Step S409: After completing the processing, the maintenance personnel write back the maintenance records, fault handling results, component replacement records, and spare parts consumption information to the life assessment process to provide data input for the next round of life calculation, forming a post-maintenance feedback loop.

[0012] The multi-source data in step S4 includes maintenance plan, maintenance execution record, fault work order, self-inspection record, spare parts inventory flow, production line location, equipment start-up time, equipment status, and manual remarks. The lifetime feature vector X is constructed using the following formula:

[0013] in, This refers to the current cumulative usage time or equivalent usage time. This represents the current maintenance cycle consumption rate. The number of overdue days Number of failures This represents the number of self-check anomalies. This is the most recent maintenance quality rating. This is a spare parts replacement status. To mitigate the impact of advance warning, For equipment status or production line importance level, To identify anomalous semantic risks inferred by large AI models from text records; The lifetime decay index D is calculated using the normalization function Normalize, and its formula is as follows:

[0014] in, to For feature weights, , , , , , , for , , , , , , Normalized eigenvalues An additional attenuation coefficient is added to characterize the accelerated attenuation of component lifespan caused by cycle consumption, overdue penalties, fault penalties, self-inspection anomalies, maintenance quality, equipment importance, and AI semantic risks. The higher the value, the closer the component is to failure. The remaining useful life (RUL) is calculated using the following formula:

[0015] in, For the reference life of the component, For equivalent usage time, For the additional attenuation coefficient, The lifespan reset amount resulting from effective maintenance. This is the life reset amount resulting from spare parts or component replacement. When a component or spare part is replaced, the system can... Reset to zero or reduce by replacement ratio, and put the component into a new life cycle; The Health Index (HI) is calculated using the following formula:

[0016] in, For remaining lifespan, This is the reference life of the component.

[0017] The AI ​​large model decision layer in step S5 includes the following AI large model decision steps: Step S501: Read the life assessment results, fault reporting information, historical maintenance records, and spare parts inventory information to assemble the context. The life assessment results include the remaining life (RUL), health index (HI), and risk level. The fault reporting information includes the title, details, fault level, and on-site photos of the fault report. The historical maintenance records include the completion time, processing results, and cycle attribution of the historical maintenance records. The spare parts inventory information includes spare parts inventory, spare parts ledger, safety stock, and inventory flow. Step S502: When the risk level reaches the level of attention, warning, abnormality, or when maintenance personnel submit a fault work order, the equipment name, equipment number, production line, component name, module name, self-inspection abnormality, historical maintenance records, spare parts inventory information, life status, life assessment results, fault repair information, current maintenance plan, and standard maintenance methods in the equipment template are organized into structured prompt words and input into the AI ​​big model decision engine. Step S503: The AI ​​big model decision engine performs semantic understanding of the fault, and outputs a candidate list of fault causes, inspection steps, repair strategies, maintenance strategies, spare parts suggestions, downtime suggestions, and cycle suggestions by combining the knowledge base and historical records. The repair strategy includes repair steps, responsible person suggestions, and risk explanations, and is pushed to the fault work order or maintenance task for on-site execution. The spare parts suggestions generate recommended spare parts, spare parts warnings, and inventory replenishment suggestions based on the repair strategy and inventory information, and prepare the materials required for repair in advance. The cycle suggestions generate suggestions to shorten, maintain, or extend the cycle based on the fault frequency, life assessment results, repair results, and AI judgment, and pass the suggestions to the maintenance cycle rolling update module. Step S504: On-site execution and confirmation. Maintenance personnel handle the maintenance according to the maintenance strategy output in step S503, and fill in the processing results, actual spare parts used, downtime status and manual confirmation opinions to the historical maintenance records, spare parts inventory and knowledge base, and feed them back to the AI ​​big model decision engine.

[0018] The periodic rolling update layer in step S6 includes the following maintenance periodic rolling update steps: Step S601: Submit maintenance record. After completing maintenance, troubleshooting or component replacement, the maintenance personnel shall submit a maintenance record, which shall record the equipment instance, maintenance item, component name, handling result, actual completion time and operator. Step S602: Read the actual completion time. Read the actual maintenance completion time in the maintenance record and use it as the time base for the current cycle rolling calculation. Step S603: Determine whether spare parts or critical components need to be replaced in this process based on maintenance records, spare parts outbound logs, and component replacement records; Step S604: Calculate the life recovery amount. If a critical component is replaced, the life of the component is fully reset or partially restored according to the replacement ratio. If only cleaning, adjustment or routine maintenance is completed, the partial life recovery amount is calculated based on the maintenance quality and treatment results. If the treatment is ineffective, the life recovery amount is not increased. Step S605: Recalculate the component life status, and recalculate the remaining life (RUL), health index (HI), and risk level by taking into account the life recovery amount, number of failures, self-test anomalies, overdue status, and usage time. Step S606: Calculate the CycleFactor. The CycleFactor is determined based on the Health Index (HI), failure frequency, self-test anomalies, overdue status, component importance level, and AI cycle recommendations. When the health status is good, the CycleFactor can remain unchanged or be appropriately increased. When there is a high risk, the CycleFactor is decreased to shorten the next maintenance cycle. Step S607: Generate the next maintenance due time. Calculate the adjusted maintenance interval based on the actual maintenance completion time formula, and generate the next maintenance due time based on the next maintenance time formula. The formula for the actual maintenance completion time is as follows:

[0019] in, Basic maintenance cycle, The periodic adjustment factor is jointly determined by the health index, failure frequency, self-check anomalies, and AI periodic suggestions. The formula for the next maintenance time is as follows: ; Step S608: Write the cycle attribution. Write the cycle due date (CycleDueDate) corresponding to this maintenance into the maintenance record, and use a combination of "equipment + maintenance item + cycle due date", "equipment + component + maintenance item + cycle number" or "plan number + cycle attribution" for unique verification. Step S609: Update the maintenance plan and life objects, and write the next maintenance due time, cycle attribution, life recovery result and risk level generated in step S607 into the maintenance plan, component life digital twin object and production line dashboard; Step S610: Continuous feedback is generated, and subsequent maintenance, fault reporting, self-inspection anomalies and spare parts replacement records continue to be written back to the life model and fed back to step S602.

[0020] The application presentation layer in step S7 includes the following production line dashboard status aggregation steps: Step S701: Read the production line and equipment layout data, which includes the production line background image, equipment coordinates X / Y, main line equipment identifiers and other equipment identifiers, and determine the display position and equipment type of the equipment in the production line Kanban. Step S702: Read equipment status data, which includes status data, self-inspection records, maintenance plans, component replacement records, and maintenance execution records, serving as the data source for production line dashboard status aggregation; Step S703: Call the production line Kanban service, and through the production line Kanban service call the layout reading interface, status reading interface, self-test reading interface, component lifespan interface and maintenance progress interface to obtain layout information, equipment status, self-test status, component lifespan status and maintenance progress respectively; Step S704: Perform status aggregation. Generate equipment aggregate status based on self-inspection results, component life status, and maintenance progress. The self-inspection results include normal and abnormal, the component life status includes normal, warning, and abnormal, and the maintenance progress status includes progress percentage. Step S705: Generate the main line equipment workstation status. Based on the X / Y positioning information of the main line equipment, map the main line equipment to the workstation unit in the front-end dashboard, and display the equipment status, component life status and maintenance progress in the workstation unit. Step S706: Generate other equipment status. Generate other equipment status based on the component lifespan and maintenance progress of auxiliary equipment or non-mainline equipment, and display the mainline equipment and other equipment in the front-end dashboard. Step S707: Output the front-end dashboard display results, outputting the production line background image, equipment markers, status colors, and details pop-ups to the front-end dashboard; Step S708: Determine the status color. Based on the aggregated device status, determine the front-end display color. "ok" indicates normal operation and is displayed in green; "warning" indicates a warning and is displayed in yellow; "danger" indicates an abnormality and is displayed in red; "missing" indicates no self-test and is displayed in gray; "disabled" indicates disabled or pending activation and is displayed in blue.

[0021] Compared with the prior art, the beneficial effects of the present invention are: (1) Maintenance management granularity is reduced to the core components of equipment. The component life digital twin object is used as an independent calculation unit to realize refined component-level predictive maintenance and accurately locate high-wear faulty components. (2) Integrate maintenance, fault, self-inspection, spare parts, and text semantic multi-source data to construct life feature vector, quantify the component wear rate through decay index, and intuitively reflect the real health status with remaining life RUL and health index HI, thoroughly solve the problem of under-warranty and over-warranty caused by fixed cycle, and reduce downtime failure and spare parts waste; (3) The rule-based lifespan algorithm works in conjunction with the AI ​​big model to automatically output standardized maintenance steps, spare parts lists, and root causes of failures, reducing reliance on the experience of senior maintenance personnel and simultaneously accumulating enterprise maintenance knowledge base; (4) Adaptive rolling update of maintenance cycle, dynamically adjust maintenance interval based on actual maintenance completion time, automatically correct life model after early / late maintenance and component replacement, and eliminate cycle drift problem; (5) The maintenance cycle is uniquely verified by the maintenance period due time, avoiding the duplicate creation of maintenance work orders for the same period and standardizing operation and maintenance business data; (6) The component life risk is deeply linked with spare parts inventory and production line dashboard, which provides early warning of spare parts shortage risk. The visual dashboard uses color blocks to intuitively display the health status of the entire line equipment components, allowing managers to quickly identify potential problems. (7) Establish a complete closed loop of “data collection - life quantification assessment - AI intelligent maintenance decision - adaptive cycle scheduling - on-site maintenance execution - data write-back optimization”. The system continuously optimizes itself as maintenance data accumulates, and adapts to different production lines and different types of equipment operating conditions. (8) It is compatible with the existing traditional maintenance system data table structure, with low transformation and implementation costs. It can be smoothly connected to the existing equipment ledger, work order and inventory modules without large-scale reconstruction of the original business system.

[0022] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the present invention more apparent and understandable, the following lists specific implementation methods of the present invention.

[0023] The above and other objects, features and advantages of the present invention will become more apparent to those skilled in the art from the following detailed description of specific embodiments of the invention in conjunction with the accompanying drawings, but this is not intended to limit the invention. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the structure of the present invention; Figure 2 This is a schematic diagram of the component life digital twin model of the present invention; Figure 3 This is a schematic diagram of the component life assessment of the present invention; Figure 4 This is a schematic diagram of the AI ​​large-scale model decision-making of the present invention; Figure 5 This is a schematic diagram of the rolling update of the maintenance cycle of the present invention; Figure 6 This is a schematic diagram of the production line Kanban status aggregation of the present invention. Detailed Implementation

[0025] To facilitate understanding of the present invention, a more comprehensive description will be provided below. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the present invention.

[0026] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0027] The specific embodiments provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0028] like Figure 1As shown, a real-time data-driven AI-powered SmartTPM predictive maintenance system includes a data source layer, a WebAPI service layer, a database entity layer, a component life assessment layer, an AI large model decision layer, an application display layer, a periodic rolling update layer, and an execution feedback closed loop. A method for implementing a real-time data-driven, AI-based SmartTPM predictive maintenance system includes the following steps: Step S1: The data source layer provides basic data to the WebAPI service layer through business calls. The basic data includes equipment templates, modules, components, maintenance items, equipment instances, production line locations, self-inspection records, fault work orders, spare parts ledgers, and inventory flow. Step S2: The WebAPI service layer receives business calls through the equipment management service, production line dashboard service, fault reporting service, and spare parts inventory service, and provides business data to the component life assessment layer through aggregation calculation; Step S3: The database entity layer stores data and provides data support for the component life assessment layer. The stored data includes maintenance plan, maintenance execution record, self-inspection record, fault work order, and spare parts inventory flow. Step S4: The component life assessment layer constructs a life feature vector X based on multi-source data, calculates the life decay index D, remaining life RUL, health index HI and risk level, and outputs the status to the application display layer, and outputs the life status and risk context to the AI ​​big model decision layer. Step S5: The AI ​​big model decision layer generates maintenance steps, spare parts suggestions, risk explanations and cycle adjustment suggestions based on fault semantic understanding, historical maintenance records, component life status, spare parts inventory and knowledge base, and outputs the strategy suggestions to the application display layer and the cycle suggestions to the cycle rolling update layer. Step S6: The cycle rolling update layer dynamically calculates the cycle adjustment factor CycleFactor, the next maintenance due time, and the cycle assignment based on the actual maintenance completion time, component life assessment results, and cycle recommendations, and outputs the update task to the application display layer. Step S7: The application display layer will output the status received from the component life assessment layer to the personal workbench, data overview, equipment management, fault reporting, spare parts inventory and production line dashboard, and output work assignment / display to the execution feedback loop; Step S8: Execute the feedback loop to write back the records to the database entity layer of step S3. The content of the record write-back includes maintenance completion, fault handling, component replacement, and inventory in / out results.

[0029] The data source layer in step S1 includes the following steps: Step S101: Establish the equipment template layer structure. The equipment template layer includes equipment templates, modules, components, and maintenance items. The equipment template describes the standard structure of a certain type of equipment. The module divides the internal functional areas of the equipment. The component records the component name, component baseline life L0, and component importance level of the maintainable object. The maintenance item records the maintenance content, maintenance cycle, and early warning lead time of the corresponding component. For example, under the glue spraying equipment template, a glue spraying module, a conveying module, and an electrical control module can be set. The glue spraying module includes components such as nozzles, cylinders, and sensors. Each component can be bound to one or more maintenance items. Step S102: Configure baseline life parameters for each component. The baseline life parameters include design life days, default maintenance cycle, component importance level, early warning lead time, and optional spare parts replacement strategy. Step S103: Create an equipment instance, bind the equipment instance to the production line location and equipment template, and map the template component to a component life digital twin object under the equipment instance. The component life digital twin object includes the following fields: component identifier, equipment identifier, module name, component name, component baseline life L0, number of days used or equivalent usage time U, last maintenance completion time Tm, next maintenance due time Td, number of failures F, number of self-inspection anomalies Q, overdue days O, spare parts replacement status P, health index HI, remaining life RUL, and risk level RiskLevel. The component's baseline lifespan L0 can be determined by one or more of the following methods: component configuration lifespan, shortest maintenance cycle, weighted maintenance cycle, historical fault statistics, equipment manufacturer's recommended lifespan, or human experience lifespan. RiskLevel includes Normal, Watchlist, Warning, Abnormal, and Frozen states. Normal state indicates that the component health index (HI) is high and has not entered the warning window; Watchlist state indicates that the remaining useful life (RUL) is decreasing or about to enter the warning window; Warning state indicates that there are multiple abnormalities, overdue, or a significant decrease in the health index (HI); Abnormal state indicates that the remaining useful life (RUL) is zero, a serious failure has occurred, or immediate repair is required; Frozen state indicates that the equipment is out of service, not activated, or temporarily not included in the life calculation. Step S104: As Figure 2 As shown, a digital twin model of component lifespan is constructed, including an equipment template layer, an equipment instance layer, a maintenance plan snapshot layer, a component lifespan status layer, a component replacement record layer, and a dashboard display layer. The equipment instance layer records the equipment actually put into production, including equipment number, production line, activation time and Kanban coordinates. When an equipment instance is bound to an equipment template, a corresponding maintenance plan snapshot is automatically generated based on the modules, components and maintenance items in the template. The maintenance plan snapshot saves the maintenance status of the equipment instance within the current cycle. The maintenance status includes equipment identifier, module name, component name, last completion time, next maintenance due time, cycle maintenance due time, and maintenance item content. The maintenance plan snapshot serves as both a data source for lifespan calculation and a basis for generating maintenance tasks and determining cycle attribution. The component life status layer is generated from the component life digital twin object with "equipment instance + component + maintenance plan" as the basic unit. It stores the component's baseline life L0, the number of days used or equivalent usage time U, the time of the most recent maintenance completion Tm, the time of the next maintenance due Td, the number of failures F, the number of self-inspection anomalies Q, the number of overdue days O, the spare parts replacement status P, the health index HI, the remaining life RUL, and the risk level RiskLevel. The component replacement record layer records the replacement time, replaced component, spare parts information, operator and remarks. When the system identifies that a critical component has been replaced, it resets or partially restores the component's lifespan based on the replacement record and recalculates the next maintenance time Td. The dashboard display layer shows a list of component lifespans, remaining days, health index (HI), and status color based on the component lifespan digital twin object.

[0030] In step S103, the component baseline lifespan L0 is initialized. If the component is configured with a component lifespan in days (ComponentLifeDays), then the component baseline lifespan L0 = ComponentLifeDays. If there is a maintenance item cycle (IntervalDays) under the component, then L0 = the shortest critical maintenance cycle or the weighted maintenance cycle. Otherwise, L0 = the system default lifespan parameter. When multiple maintenance items apply to the same component simultaneously, a weighted fusion formula can be used:

[0031] in, , , For weight parameters, This is the shortest cycle among the maintenance items related to the component. The average cycle of maintenance items related to the component is given by CriticalLevelFactor, which is a correction factor obtained from Level 1 maintenance, Level 2 maintenance, Level 3 maintenance, or component importance level.

[0032] The component life assessment layer in step S4 includes the following component life assessment steps, such as... Figure 3 As shown: Step S401: Read the equipment instance and template structure, read the hierarchical relationship between equipment instances, equipment templates, modules, components and maintenance items, and determine the equipment and component objects to be assessed for life. Step S402: Summarize the maintenance plan snapshot and read the maintenance plan information corresponding to the component, including the number of days of the maintenance cycle, the next maintenance due date, the warning lead time, the last maintenance completion time and the cycle maintenance due date; Step S403: Read the component lifespan start point. Determine the start time for component lifespan calculation based on the equipment activation time or the most recent component replacement time. If a component replacement record exists, use the replacement time as the new lifespan cycle start point. Step S404: Calculate the lifespan baseline and the number of days used. Determine the component baseline lifespan L0 based on the component lifespan days, maintenance item cycle, or replacement record lifespan snapshot, and calculate the number of days used or equivalent usage time U based on the current date, activation time, maintenance records, and replacement records. Step S405: Construct a lifespan feature vector X based on the number of days used or equivalent usage time U, the current maintenance cycle consumption rate C, the number of overdue days O, the number of failures F, the number of self-inspection anomalies Q, the most recent maintenance quality score M, the spare parts replacement status P, the impact of early warning lead time W, the equipment status or production line importance level S, and the abnormal semantic risk A inferred by the AI ​​big model from the text records. Step S406: Calculate the additional attenuation index D based on cycle consumption, overdue penalty, fault penalty, self-inspection anomaly, maintenance quality, equipment importance and AI semantic risk factors; Step S407: Based on the component's reference life L0 and equivalent service life Life decay index D, life recovery amount due to effective maintenance And the life reset amount caused by spare parts replacement or component replacement Calculate the remaining life expectancy (RUL) and convert it into a health index (HI), while also outputting the risk level. Step S408: Determine the risk level when... And remaining lifespan A value greater than the warning window is considered normal, and the current maintenance cycle should be maintained. or remaining lifespan Enter the warning window as a watchlist. Multiple anomalies may trigger an early warning, providing advance reminders or shortening the maintenance cycle. And remaining lifespan A serious malfunction may be detected as an anomaly, triggering a repair / maintenance request, shutdown for inspection, or spare parts preparation. If the equipment is out of service or not effective, it will be frozen. Step S409: After completing the processing, the maintenance personnel write back the maintenance records, fault handling results, component replacement records, and spare parts consumption information to the life assessment process to provide data input for the next round of life calculation, forming a post-maintenance feedback loop.

[0033] The multi-source data in step S4 includes maintenance plan, maintenance execution record, fault work order, self-inspection record, spare parts inventory flow, production line location, equipment start-up time, equipment status, and manual remarks; The lifetime feature vector X is constructed using the following formula:

[0034] in, This refers to the current cumulative usage time or equivalent number of days. This represents the current maintenance cycle consumption rate. The number of overdue days Number of failures This represents the number of self-check anomalies. This is the most recent maintenance quality rating. This is a spare parts replacement status. To mitigate the impact of advance warning, For equipment status or production line importance level, For abnormal semantic risks inferred by AI large models from text records, the AI ​​large model can be a cloud-based large model, a local private model, an industry knowledge base-enhanced model, or a rule template can be used to simulate and output maintenance suggestions before replacing them with a real large model. The composition of the life feature vector X is not limited to the above fields. It can be expanded, deleted or combined according to the equipment type, production environment and data availability. Any scheme that constructs life assessment features based on one or more of the following data: component usage status, maintenance records, fault records, self-inspection anomalies, spare parts replacement, equipment status or AI semantic risks, and is used to calculate the remaining life, health index or risk level of the component, is a component field of the life feature vector X. The lifetime decay index D is calculated using the normalization function Normalize, and its formula is as follows:

[0035] in, to For feature weights, , , , , , , for , , , , , , Normalized eigenvalues An additional attenuation coefficient is added to characterize the accelerated attenuation of component lifespan caused by cycle consumption, overdue penalties, fault penalties, self-inspection anomalies, maintenance quality, equipment importance, and AI semantic risks. The higher the value, the closer the component is to failure. Feature Term Not directly involved Instead of weighted calculation, it is used as Participated in the calculation of remaining lifetime (RUL); characteristic term Not directly involved It is not a weighted calculation, but is used to determine whether a weighted calculation has occurred. Or whether to reset the life cycle; Feature item Not directly involved The weighted calculation is not used to determine the early warning window and early alert threshold when assessing risk levels; The remaining useful life (RUL) is calculated using the following formula:

[0036] in, For the reference life of the component, The equivalent usage time can be calculated from the activation time, operating load, maintenance records, fault records, and component replacement records. An additional attenuation factor is added to amplify the impact of equivalent usage time on lifetime depletion, rather than directly representing the proportion of lifetime already consumed. The lifespan reset amount resulting from effective maintenance. This is the life reset amount resulting from spare parts or component replacement. When a component or spare part is replaced, the system can... Reset to zero or reduce by replacement ratio, and put the component into a new life cycle; The Health Index (HI) is calculated using the following formula:

[0037] Among them, the closer the health index HI is to 100, the healthier the component; the closer it is to 0, the closer the component is to failure. For remaining lifespan, This is the reference life of the component; The Health Index (HI) can be calculated as the ratio of Remaining Life (RUL) to Part Baseline Life (L0), or it can be calculated by reverse mapping of risk scores, piecewise functions, rating scales, or model output probabilities.

[0038] The AI ​​large model decision layer in step S5 includes the following AI large model decision steps, such as... Figure 4 As shown: Step S501: Read the life assessment results, fault reporting information, historical maintenance records, and spare parts inventory information to assemble the context. The life assessment results include the remaining life (RUL), health index (HI), and risk level. The fault reporting information includes the title, details, fault level, and on-site photos of the fault report. The historical maintenance records include the completion time, processing results, and cycle attribution of the historical maintenance records. The spare parts inventory information includes spare parts inventory, spare parts ledger, safety stock, and inventory flow. Step S502: When the risk level reaches the level of attention, warning, abnormality, or when maintenance personnel submit a fault work order, the equipment name, equipment number, production line, component name, module name, self-inspection abnormality, historical maintenance records, spare parts inventory information, life status, life assessment results, fault repair information, current maintenance plan, and standard maintenance methods in the equipment template are organized into structured prompt words and input into the AI ​​big model decision engine. Step S503: The AI ​​big model decision engine performs semantic understanding of the fault, and outputs a candidate list of fault causes, inspection steps, repair strategies, maintenance strategies, spare parts suggestions, downtime suggestions, and cycle suggestions by combining the knowledge base and historical records. The repair strategy includes repair steps, responsible person suggestions, and risk explanations, and is pushed to the fault work order or maintenance task for on-site execution. The spare parts suggestions generate recommended spare parts, spare parts warnings, and inventory replenishment suggestions based on the repair strategy and inventory information, and prepare the materials required for repair in advance. The cycle suggestions generate suggestions to shorten, maintain, or extend the cycle based on the fault frequency, life assessment results, repair results, and AI judgment, and pass the suggestions to the maintenance cycle rolling update module. Step S504: On-site execution and confirmation. Maintenance personnel shall handle the maintenance according to the maintenance strategy output in step S503, and fill in the processing results, actual spare parts used, downtime status and manual confirmation opinions to the historical maintenance records, spare parts inventory and knowledge base, and feed them back to the AI ​​big model decision engine. The rule algorithm is responsible for determining "whether the component is healthy, how long it can be used, and whether a task should be triggered", while the AI ​​big model is responsible for explaining "why, how to repair, what spare parts to use, and whether the cycle needs to be adjusted". When the two work together, the system can combine quantitative judgment of lifespan with on-site maintenance strategies.

[0039] The periodic rolling update layer in step S6 includes the following maintenance periodic rolling update steps, such as... Figure 5 As shown: Step S601: Submit maintenance record. After completing maintenance, troubleshooting or component replacement, the maintenance personnel shall submit a maintenance record, which shall record the equipment instance, maintenance item, component name, handling result, actual completion time and operator. Step S602: Read the actual completion time. Read the actual maintenance completion time in the maintenance record and use it as the time base for the current cycle rolling calculation, instead of simply using the original planned maintenance time. Step S603: Determine whether spare parts or critical components need to be replaced in this process based on maintenance records, spare parts outbound logs, and component replacement records; Step S604: Calculate the life recovery amount. If a critical component is replaced, the life of the component is fully reset or partially restored according to the replacement ratio. If only cleaning, adjustment or routine maintenance is completed, the partial life recovery amount is calculated based on the maintenance quality and treatment results. If the treatment is ineffective, the life recovery amount is not increased. Step S605: Recalculate the component life status, and recalculate the remaining life (RUL), health index (HI), and risk level by taking into account the life recovery amount, number of failures, self-test anomalies, overdue status, and usage time. Step S606: Calculate the CycleFactor. The CycleFactor is determined based on the Health Index (HI), failure frequency, self-test anomalies, overdue status, component importance level, and AI cycle recommendations. When the health status is good, the CycleFactor can remain unchanged or be appropriately increased. When there is a high risk, the CycleFactor is decreased to shorten the next maintenance cycle. Step S607: Generate the next maintenance due time. Calculate the adjusted maintenance interval based on the actual maintenance completion time formula, and generate the next maintenance due time based on the next maintenance time formula. The formula for actual maintenance completion time is:

[0040] in, Basic maintenance cycle, The periodic adjustment factor is jointly determined by the health index, failure frequency, self-check anomalies, and AI periodic suggestions. The formula for the next maintenance time is: ; If the health index HI is high and there are no abnormalities, the cycle can be maintained or extended appropriately. If the health index HI is moderate or enters the warning window, the cycle can be shortened. If the health index HI is low or there is a serious fault, repair or replacement can be triggered immediately. If the replacement of spare parts is completed, the life cycle can be re-initialized. The maintenance cycle can be updated on a rolling basis, either by extending the cycle according to the actual completion time or by a combination of factors such as the maintenance due time, downtime window, production line scheduling, spare parts delivery time, or risk level. Step S608: Write the cycle attribution. Write the CycleDueDate corresponding to this maintenance into the maintenance record, and use a combination of "equipment + maintenance item + cycle due date", "equipment + component + maintenance item + cycle number" or "plan number + cycle attribution" for unique verification to prevent duplicate submissions in the same cycle. In scenarios where maintenance needs to be accurate to the component level, the latter combination is preferred to avoid confusion between maintenance records of different components under the same equipment. Step S609: Update the maintenance plan and life objects, and write the next maintenance due time, cycle attribution, life recovery result and risk level generated in step S607 into the maintenance plan, component life digital twin object and production line dashboard; Step S610: A rolling feedback is generated. Subsequent maintenance, fault reporting, self-inspection anomalies, and spare parts replacement records are continued to be written back to the life model and fed back to step S602 for the next cycle of rolling updates.

[0041] The application presentation layer in step S7 includes the following production line Kanban status aggregation steps, such as... Figure 6 As shown: Step S701: Read the production line and equipment layout data, which includes the production line background image, equipment coordinates X / Y, main line equipment identifiers and other equipment identifiers, and determine the display position and equipment type of the equipment in the production line Kanban. Step S702: Read equipment status data, which includes status data, self-inspection records, maintenance plans, component replacement records, and maintenance execution records, serving as the data source for production line dashboard status aggregation; Step S703: Call the production line Kanban service, and through the production line Kanban service call the layout reading interface, status reading interface, self-test reading interface, component lifespan interface and maintenance progress interface to obtain layout information, equipment status, self-test status, component lifespan status and maintenance progress respectively; Step S704: Perform status aggregation. Generate equipment aggregate status based on self-inspection results, component life status, and maintenance progress. The self-inspection results include normal and abnormal, the component life status includes normal, warning, and abnormal, and the maintenance progress includes progress percentage. Step S705: Generate the main line equipment workstation status. Based on the X / Y positioning information of the main line equipment, map the main line equipment to the workstation unit in the front-end dashboard, and display the equipment status, component life status and maintenance progress in the workstation unit. Step S706: Generate other equipment status. Generate other equipment status based on the component lifespan and maintenance progress of auxiliary equipment or non-mainline equipment, and display the mainline equipment and other equipment in the front-end dashboard. Step S707: Output the front-end dashboard display results. Output the production line background image, equipment markings, status colors, and details pop-ups to the front-end dashboard so that managers can view equipment location, status color, self-inspection status, component life status, maintenance progress, and equipment details. Step S708: Determine the status color. Based on the aggregated device status, determine the front-end display color. "ok" indicates normal operation and is displayed in green; "warning" indicates a warning and is displayed in yellow; "danger" indicates an abnormality and is displayed in red; "missing" indicates no self-test and is displayed in gray; "disabled" indicates disabled or pending activation and is displayed in blue.

[0042] Kanban displays can be aggregated by equipment, production line, module, component, risk level, spare parts shortage, or maintenance manager.

[0043] Example 1 Taking a glue spraying equipment on a production line as an example, the equipment template includes a glue spraying module, a conveying module, and an electrical control module; the glue spraying module includes components such as nozzles, cylinders, and sensors. The system is configured with a baseline lifespan of 60 days for the nozzles and 180 days for the cylinders, with different maintenance cycles and early warning lead times set for each.

[0044] After device A001 is put into use on June 1, 2026, the system generates digital twin objects of the lifespan of components such as nozzles, cylinders, and sensors for A001 based on the template. The system updates the equivalent number of days of use, fault penalty factor and self-inspection anomaly factor of each component every day based on the current date, maintenance plan, fault work order and self-inspection record.

[0045] If the nozzle component exhibits two self-inspection anomalies by June 20, 2026, and the next maintenance due date has already entered the early warning window, the system calculates that the nozzle's health index drops to 42, and the risk level is warning. The system inputs the nozzle's lifespan status, anomaly description, and historical maintenance records into the AI ​​big model. The big model outputs maintenance suggestions such as "check nozzle blockage, clean adhesive lines, check air pressure stability, and prepare nozzle spare parts," and recommends shortening the next cycle from 7 days to 5 days.

[0046] After on-site personnel complete the cleaning and submit the maintenance record, the system updates the last maintenance time based on the actual completion time and partially restores the nozzle lifespan based on the processing results. If nozzle spare parts are also registered for replacement, the system resets the nozzle life cycle and recalculates the next maintenance due date. In the production line dashboard, the component lifespan status of the equipment changes from abnormal or warning to normal, and the remaining nozzle lifespan, health index, next maintenance due date, AI cycle suggestion, and spare parts consumption record are displayed in the equipment details.

[0047] The technical features described in the above examples can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above examples are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0048] The examples described above merely illustrate embodiments of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and non-substantial improvements without departing from the concept of the present invention, and these all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A real-time data-driven artificial intelligence SmartTPM predictive maintenance system, characterized in that, It includes a data source layer, a WebAPI service layer, a database entity layer, a component life assessment layer, an AI large model decision layer, an application display layer, a periodic rolling update layer, and an execution feedback closed loop; A method for implementing a real-time data-driven, AI-based SmartTPM predictive maintenance system includes the following steps: Step S1: The data source layer provides basic data to the WebAPI service layer through business calls. The basic data includes equipment templates, modules, components, maintenance items, equipment instances, production line locations, self-inspection records, fault work orders, spare parts ledgers, and inventory flow. Step S2: The WebAPI service layer receives business calls through the equipment management service, production line dashboard service, fault reporting service, and spare parts inventory service, and provides business data to the component life assessment layer through aggregation calculation; Step S3: The database entity layer stores data and provides data support for the component life assessment layer. The stored data includes maintenance plan, maintenance execution record, self-inspection record, fault work order, and spare parts inventory flow. Step S4: The component life assessment layer constructs a life feature vector X based on multi-source data, calculates the life decay index D, remaining life RUL, health index HI and risk level, and outputs the status to the application display layer, and outputs the life status and risk context to the AI ​​big model decision layer. Step S5: The AI ​​big model decision layer generates maintenance steps, spare parts suggestions, risk explanations and cycle adjustment suggestions based on fault semantic understanding, historical maintenance records, component life status, spare parts inventory and knowledge base, and outputs the strategy suggestions to the application display layer and the cycle suggestions to the cycle rolling update layer. Step S6: The cycle rolling update layer dynamically calculates the cycle adjustment factor CycleFactor, the next maintenance due time, and the cycle assignment based on the actual maintenance completion time, component life assessment results, and cycle recommendations, and outputs the update task to the application display layer. Step S7: The application display layer will output the status received from the component life assessment layer to the personal workbench, data overview, equipment management, fault reporting, spare parts inventory and production line dashboard, and output work assignment / display to the execution feedback loop; Step S8: Execute the feedback loop to write back the records to the database entity layer of step S3. The content of the record write-back includes maintenance completion, fault handling, component replacement, and inventory in / out results.

2. The real-time data-driven artificial intelligence SmartTPM predictive maintenance system according to claim 1, characterized in that, The data source layer in step S1 includes the following steps: Step S101: Establish the equipment template layer structure. The equipment template layer includes equipment templates, modules, components, and maintenance items. The equipment template describes the standard structure of a certain type of equipment. The module divides the internal functional areas of the equipment. The component records the component name, component baseline life L0, and component importance level of the maintainable object. The maintenance item records the maintenance content, maintenance cycle, and early warning lead time of the corresponding component. Step S102: Configure baseline life parameters for each component. The baseline life parameters include design life days, default maintenance cycle, component importance level, early warning lead time, and optional spare parts replacement strategy. Step S103: Create an equipment instance, bind the equipment instance to the production line location and equipment template, and map the template component to a component life digital twin object under the equipment instance. The component life digital twin object includes the following fields: component identifier, equipment identifier, module name, component name, component baseline life L0, number of days used or equivalent usage time U, last maintenance completion time Tm, next maintenance due time Td, number of failures F, number of self-inspection anomalies Q, overdue days O, spare parts replacement status P, health index HI, remaining life RUL, and risk level RiskLevel. The RiskLevel includes Normal, Attention, Warning, Abnormal, and Frozen status; Step S104: Construct a digital twin model of component lifespan, including an equipment template layer, an equipment instance layer, a maintenance plan snapshot layer, a component lifespan status layer, a component replacement record layer, and a dashboard display layer; The equipment instance layer records the equipment actually put into production, including equipment number, production line, activation time and Kanban coordinates. When an equipment instance is bound to an equipment template, a corresponding maintenance plan snapshot is automatically generated based on the modules, components and maintenance items in the template. The maintenance plan snapshot saves the maintenance status of the device instance within the current cycle. The maintenance status includes device identifier, module name, component name, last completion time, next maintenance due time, cycle maintenance due time, and maintenance item content. The component life status layer is generated by the component life digital twin object with "equipment instance + component + maintenance plan" as the basic unit, and stores the component's baseline life L0, number of days used or equivalent usage time U, the time of the most recent maintenance completion Tm, the time of the next maintenance due Td, the number of failures F, the number of self-inspection anomalies Q, the number of overdue days O, the spare parts replacement status P, the health index HI, the remaining life RUL, and the risk level RiskLevel. The component replacement record layer records the replacement time, replaced component, spare parts information, operator and remarks. When the system identifies that a critical component has been replaced, it resets or partially restores the component's lifespan based on the replacement record and recalculates the next maintenance time Td. The aforementioned dashboard display layer displays a list of component lifespans, remaining days, health index (HI), and status color based on the component lifespan digital twin object.

3. The real-time data-driven artificial intelligence SmartTPM predictive maintenance system according to claim 2, characterized in that, In step S103, the component baseline lifespan L0 is initialized. If the component is configured with a component lifespan in days (ComponentLifeDays), then the component baseline lifespan L0 = ComponentLifeDays. If there is a maintenance item cycle (IntervalDays) under the component, then L0 = the shortest critical maintenance cycle or the weighted maintenance cycle. Otherwise, L0 = the system default lifespan parameter. When multiple maintenance items apply to the same component simultaneously, a weighted fusion formula can be used: in, , , For weight parameters, This is the shortest cycle among the maintenance items related to the component. The average cycle of maintenance items related to the component is given by CriticalLevelFactor, which is a correction factor obtained from Level 1 maintenance, Level 2 maintenance, Level 3 maintenance, or component importance level.

4. The real-time data-driven artificial intelligence SmartTPM predictive maintenance system according to claim 1, characterized in that, The component life assessment layer in step S4 includes the following component life assessment steps: Step S401: Read the equipment instance and template structure, read the hierarchical relationship between equipment instances, equipment templates, modules, components and maintenance items, and determine the equipment and component objects to be assessed for life. Step S402: Summarize the maintenance plan snapshot and read the maintenance plan information corresponding to the component, including the number of days of the maintenance cycle, the next maintenance due date, the warning lead time, the last maintenance completion time and the cycle maintenance due date; Step S403: Read the component lifespan start point. Determine the start time for component lifespan calculation based on the equipment activation time or the most recent component replacement time. If a component replacement record exists, use the replacement time as the new lifespan cycle start point. Step S404: Calculate the lifespan baseline and the number of days used. Determine the component baseline lifespan L0 based on the component lifespan days, maintenance item cycle, or replacement record lifespan snapshot, and calculate the number of days used or equivalent usage time U based on the current date, activation time, maintenance records, and replacement records. Step S405: Construct a lifespan feature vector X based on the number of days used or equivalent usage time U, the current maintenance cycle consumption rate C, the number of overdue days O, the number of failures F, the number of self-inspection anomalies Q, the most recent maintenance quality score M, the spare parts replacement status P, the impact of early warning lead time W, the equipment status or production line importance level S, and the abnormal semantic risk A inferred by the AI ​​big model from the text records. Step S406: Calculate the additional attenuation index D based on cycle consumption, overdue penalty, fault penalty, self-inspection anomaly, maintenance quality, equipment importance and AI semantic risk factors; Step S407: Based on the component's reference life L0 and equivalent service life Life decay index D, life recovery amount due to effective maintenance And the life reset amount caused by spare parts replacement or component replacement Calculate the remaining life expectancy (RUL) and convert it into a health index (HI), while also outputting the risk level. Step S408: Determine the risk level when... And remaining lifespan A value greater than the warning window is considered normal, and the current maintenance cycle should be maintained. or remaining lifespan Enter the warning window as a watchlist. Multiple anomalies may trigger an early warning, providing advance reminders or shortening the maintenance cycle. And remaining lifespan A serious malfunction may be detected as an anomaly, triggering a repair / maintenance request, shutdown for inspection, or spare parts preparation. If the equipment is out of service or not effective, it will be frozen. Step S409: After completing the processing, the maintenance personnel write back the maintenance records, fault handling results, component replacement records, and spare parts consumption information to the life assessment process to provide data input for the next round of life calculation, forming a post-maintenance feedback loop.

5. The real-time data-driven artificial intelligence SmartTPM predictive maintenance system according to claim 1, characterized in that, The multi-source data in step S4 includes maintenance plan, maintenance execution record, fault work order, self-inspection record, spare parts inventory flow, production line location, equipment start-up time, equipment status, and manual remarks. The lifetime feature vector X is constructed using the following formula: in, This refers to the current cumulative usage time or equivalent usage time. This represents the current maintenance cycle consumption rate. The number of overdue days Number of failures This represents the number of self-check anomalies. This is the most recent maintenance quality rating. This is a spare parts replacement status. To mitigate the impact of advance warning, For equipment status or production line importance level, To identify anomalous semantic risks inferred by large AI models from text records; The lifetime decay index D is calculated using the normalization function Normalize, and its formula is as follows: in, to For feature weights, , , , , , , for , , , , , , Normalized eigenvalues An additional attenuation coefficient is added to characterize the accelerated attenuation of component lifespan caused by cycle consumption, overdue penalties, fault penalties, self-inspection anomalies, maintenance quality, equipment importance, and AI semantic risks. The higher the value, the closer the component is to failure. The remaining useful life (RUL) is calculated using the following formula: in, For the reference life of the component, For equivalent usage time, For the additional attenuation coefficient, The lifespan reset amount resulting from effective maintenance. This is the life reset amount resulting from spare parts or component replacement. When a component or spare part is replaced, the system can... Reset to zero or reduce by replacement ratio, and put the component into a new life cycle; The Health Index (HI) is calculated using the following formula: in, For remaining lifespan, This is the reference life of the component.

6. The real-time data-driven artificial intelligence SmartTPM predictive maintenance system according to claim 1, characterized in that, The AI ​​large model decision layer in step S5 includes the following AI large model decision steps: Step S501: Read the life assessment results, fault reporting information, historical maintenance records, and spare parts inventory information to assemble the context. The life assessment results include the remaining life (RUL), health index (HI), and risk level. The fault reporting information includes the title, details, fault level, and on-site photos of the fault report. The historical maintenance records include the completion time, processing results, and cycle attribution of the historical maintenance records. The spare parts inventory information includes spare parts inventory, spare parts ledger, safety stock, and inventory flow. Step S502: When the risk level reaches the level of attention, warning, abnormality, or when maintenance personnel submit a fault work order, the equipment name, equipment number, production line, component name, module name, self-inspection abnormality, historical maintenance records, spare parts inventory information, life status, life assessment results, fault repair information, current maintenance plan, and standard maintenance methods in the equipment template are organized into structured prompt words and input into the AI ​​big model decision engine. Step S503: The AI ​​big model decision engine performs semantic understanding of the fault, and outputs a candidate list of fault causes, inspection steps, repair strategies, maintenance strategies, spare parts suggestions, downtime suggestions, and cycle suggestions by combining the knowledge base and historical records. The repair strategy includes repair steps, responsible person suggestions, and risk explanations, and is pushed to the fault work order or maintenance task for on-site execution. The spare parts suggestions generate recommended spare parts, spare parts warnings, and inventory replenishment suggestions based on the repair strategy and inventory information, and prepare the materials required for repair in advance. The cycle suggestions generate suggestions to shorten, maintain, or extend the cycle based on the fault frequency, life assessment results, repair results, and AI judgment, and pass the suggestions to the maintenance cycle rolling update module. Step S504: On-site execution and confirmation. Maintenance personnel handle the maintenance according to the maintenance strategy output in step S503, and fill in the processing results, actual spare parts used, downtime status and manual confirmation opinions to the historical maintenance records, spare parts inventory and knowledge base, and feed them back to the AI ​​big model decision engine.

7. The real-time data-driven artificial intelligence SmartTPM predictive maintenance system according to claim 1, characterized in that, The periodic rolling update layer in step S6 includes the following maintenance periodic rolling update steps: Step S601: Submit maintenance record. After completing maintenance, troubleshooting or component replacement, the maintenance personnel shall submit a maintenance record, which shall record the equipment instance, maintenance item, component name, handling result, actual completion time and operator. Step S602: Read the actual completion time. Read the actual maintenance completion time in the maintenance record and use it as the time base for the current cycle rolling calculation. Step S603: Determine whether spare parts or critical components need to be replaced in this process based on maintenance records, spare parts outbound logs, and component replacement records; Step S604: Calculate the life recovery amount. If a critical component is replaced, the life of the component is fully reset or partially restored according to the replacement ratio. If only cleaning, adjustment or routine maintenance is completed, the partial life recovery amount is calculated based on the maintenance quality and treatment results. If the treatment is ineffective, the life recovery amount is not increased. Step S605: Recalculate the component life status, and recalculate the remaining life (RUL), health index (HI), and risk level by taking into account the life recovery amount, number of failures, self-test anomalies, overdue status, and usage time. Step S606: Calculate the CycleFactor. The CycleFactor is determined based on the Health Index (HI), failure frequency, self-test anomalies, overdue status, component importance level, and AI cycle recommendations. When the health status is good, the CycleFactor can remain unchanged or be appropriately increased. When there is a high risk, the CycleFactor is decreased to shorten the next maintenance cycle. Step S607: Generate the next maintenance due time. Calculate the adjusted maintenance interval based on the actual maintenance completion time formula, and generate the next maintenance due time based on the next maintenance time formula. The formula for the actual maintenance completion time is as follows: in, Basic maintenance cycle, The periodic adjustment factor is jointly determined by the health index, failure frequency, self-check anomalies, and AI periodic suggestions. The formula for the next maintenance time is as follows: ； Step S608: Write the cycle attribution. Write the corresponding cycle due date (CycleDueDate) for this maintenance into the maintenance record and uniquely verify it using a combination of "equipment + maintenance item + cycle due date", "equipment + component + maintenance item + cycle number", or "plan number + cycle attribution". Step S609: Update the maintenance plan and life objects, and write the next maintenance due time, cycle attribution, life recovery result and risk level generated in step S607 into the maintenance plan, component life digital twin object and production line dashboard; Step S610: Continuous feedback is generated, and subsequent maintenance, fault reporting, self-inspection anomalies and spare parts replacement records continue to be written back to the life model and fed back to step S602.

8. The real-time data-driven artificial intelligence SmartTPM predictive maintenance system according to claim 1, characterized in that, The application presentation layer in step S7 includes the following production line dashboard status aggregation steps: Step S701: Read the production line and equipment layout data, which includes the production line background image, equipment coordinates X / Y, main line equipment identifiers and other equipment identifiers, and determine the display position and equipment type of the equipment in the production line Kanban. Step S702: Read equipment status data, which includes status data, self-inspection records, maintenance plans, component replacement records, and maintenance execution records, serving as the data source for production line dashboard status aggregation; Step S703: Call the production line Kanban service, and through the production line Kanban service call the layout reading interface, status reading interface, self-test reading interface, component lifespan interface and maintenance progress interface to obtain layout information, equipment status, self-test status, component lifespan status and maintenance progress respectively; Step S704: Perform status aggregation. Generate equipment aggregate status based on self-inspection results, component life status, and maintenance progress. The self-inspection results include normal and abnormal, the component life status includes normal, warning, and abnormal, and the maintenance progress includes progress percentage. Step S705: Generate the main line equipment workstation status. Based on the X / Y positioning information of the main line equipment, map the main line equipment to the workstation unit in the front-end dashboard, and display the equipment status, component life status and maintenance progress in the workstation unit. Step S706: Generate other equipment status. Generate other equipment status based on the component lifespan and maintenance progress of auxiliary equipment or non-mainline equipment, and display the mainline equipment and other equipment in the front-end dashboard. Step S707: Output the front-end dashboard display results, outputting the production line background image, equipment markers, status colors, and details pop-ups to the front-end dashboard; Step S708: Determine the status color. Based on the aggregated device status, determine the front-end display color. "ok" indicates normal operation and is displayed in green; "warning" indicates a warning and is displayed in yellow; "danger" indicates an abnormality and is displayed in red; "missing" indicates no self-test and is displayed in gray; "disabled" indicates disabled or pending activation and is displayed in blue.

Images