Digital Twin-Based Intelligent Transformation of an Igniter Production Line ()
1. Introduction
Igniter manufacturing imposes stringent requirements on both quality assurance and safety control. Owing to the discrete manufacturing characteristics of multi-variety, small-batch, and multi-batch production [1], complex routing, numerous specialized devices, and the coexistence of manual and automated operations [2] [3], its intelligent transformation faces several major challenges: fragmented production information and insufficient transparency [4]; lack of real-time closed-loop planning and scheduling with disordered resource allocation [5]; limited visual monitoring of logistics and production units [6]; and inadequate quality traceability, capacity analysis, and risk assessment, leaving production lines in a passive response state. The lack of effective closed-loop interaction between the physical and information spaces has become a critical bottleneck constraining the development of intelligent manufacturing.
Digital twin technology provides an effective approach for realizing virtual-physical information interaction and integration [7]. Through bidirectional mapping and real-time interaction between physical and information spaces, it enables dynamic monitoring, diagnosis, prediction, and optimization of production factors, representing a key enabling technology for the intelligent transformation of manufacturing processes.
Accordingly, this study takes the igniter production line as the research object and digital twin as the technical approach. It aims to overcome key technologies, including digital twin model information architecture, autonomous planning and resource scheduling algorithms, and dynamic monitoring and risk analysis of critical factors [8] [9]. A digital twin-centric technical system for the intelligent transformation of production lines is constructed to achieve visualized monitoring, analysis, and optimization of all production line factors, thereby attaining key objectives such as production process transparency, autonomous production planning, visualized production line operation and maintenance, and dynamic production line improvement. This work provides an implementable and transferable technical route for the intelligent construction of similar production lines.
The novelty of this work lies in translating a general digital-twin workshop concept into an igniter-oriented production-line framework that simultaneously addresses hazardous-operation safety, strict process-quality binding, and multi-variety small-batch delivery control. Unlike generic digital twin workshop architectures, the proposed framework is organized around the specific 4M1E objects, explosive/non-explosive process boundaries, area equivalence constraints, and human-in-the-loop control requirements of igniter manufacturing.
The principal contributions of this study can be summarized as follows:
An igniter-specific production-line digital twin information architecture is established by linking the virtual model layer, data-driven layer, analysis and computation layer, and application terminal layer to concrete 4M1E production factors.
A comprehensive factor perception and data-service middleware scheme is proposed to unify personnel, equipment, material, process, environment, and logistics data under unique object identifiers and synchronized event records.
A planning and scheduling model is developed for delivery-oriented production control under hard safety, quality, resource, routing, and batch-consistency constraints.
A closed-loop interactive platform is constructed to support risk monitoring, warning, human approval, and bounded automatic execution for production-line operation and improvement.
2. Methodology
2.1. Research Object and System Architecture
The igniter assembly process is fundamentally divided into two sequential stages: non-explosive assembly and explosive-loaded assembly [10]. The former encompasses the assembly of propellant chamber components and bridge wire spot welding, while the latter comprises powder weighing, powder loading and pressing, crimping, and electrical performance testing. The digital twin-enabled intelligent production line for igniter manufacturing is constructed upon the existing software and hardware infrastructure, with an emphasis on the collaborative integration of both assembly stages [11]. The system architecture is organized around a digital twin platform and is supported by heterogeneous production information systems to enable integrated digitalized production-line management.
2.2. Research Objectives and Functional Requirements
Intelligent manufacturing is a new production paradigm that integrates new-generation information and communication technologies, such as digital twin technology, with advanced manufacturing technologies. It is characterized by self-perception, self-learning, self-decision-making, self-execution, and self-adaptation. Compared with conventional digitalization, the application of digital twin technology can provide stronger support for the intelligent transformation of the igniter production line, particularly in quality management, safety control, and production efficiency optimization.
The functional requirements of the proposed system are categorized into three dimensions: product delivery, quality supervision, and production safety.
2.2.1. Product Delivery Dimension
The igniter production process is distinguished by extended processing sequences, high process complexity, numerous specialized equipment, and the coexistence of manual and mechanical operations. Furthermore, the production mode is characterized as “multi-variety, small-batch, and multi-batch”, where production perturbations—including schedule suspensions, order insertions, and priority changes—occur frequently. The reliance on manual experience for production planning and resource scheduling imposes considerable operational difficulty and complexity.
2.2.2. Quality Supervision Dimension
The manufacturing process involves diverse automated production and inspection equipment from multiple vendors, as well as various gauges and tools. Distinct process stages impose specific quality parameter control requirements on tooling, materials, and environmental conditions. It is therefore imperative to establish binding relationships between production and inspection processes and their corresponding quality process parameters—encompassing personnel, equipment, gauges, tooling, materials, and environmental factors—to enable process control and historical traceability.
2.2.3. Production Safety Dimension
While passive safety measures have been implemented, including anti-static facilities, explosion-proof installations, personnel electrostatic protection protocols, and area equivalence limits, and while work-in-process inventory and propellant equivalence monitoring have been realized through the Warehouse Management System (WMS), active safety management remains deficient. Specifically, there is an absence of proactive monitoring for personnel protection status, equipment safety hazards, hazardous chemical control, and environmental parameter thresholds.
To achieve the strategic objectives of production process transparency, autonomous production planning, visualized production line operation and maintenance, and dynamic production line improvement, the targeted enhancements enabled by digital twin technology across the three aforementioned dimensions are summarized in Table 1.
As indicated in Table 1, the construction of the digital twin-enabled igniter production line can be summarized as follows. Based on the existing digitalization and automation infrastructure, an ideal information model and key digital twin application technologies are developed. Production planning serves as the core driving element, while production disturbances and real-time production progress serve as triggering events. Data analysis services provide the operational logic for decision support. Through this mechanism, data twinning and virtual-physical mapping between the physical production line and its virtual counterpart are realized. Consequently, real-time perception, rapid maintenance, and dynamic optimization of production plans and production factors across the production line can be achieved.
3. Ideal Model Building
The ideal model for the igniter production line, constructed with the objective of production-line-level digital twin application, primarily comprises the information architecture model, production factor model, process layout model, manufacturing process model, and data application model [12]-[14].
Table 1. Targeted enhancements enabled by digital twin technology for the igniter production line.
Objective |
Dimension |
Objective Enhancement Content |
Production process transparency |
Product delivery |
1) Real-time monitoring of production schedule progress and real-time collection of production process data |
Quality supervision |
1) Automatic online reading of data conforming to the quality of key production process data; 2) Real-time monitoring of process consistency for personnel, equipment, and tooling; 3) Real-time judgment of the exception status of key production materials, equipment, and tooling fixtures; 4) Real-time updating of environmental temperature and humidity, key equipment operating parameters, and energy consumption data |
Production safety |
1) Online identification of personnel in blasting areas, personnel entering and leaving the operating area, and regional intrusion; 2) Real-time positioning of personnel carrying propellant products and AGV route tracking records |
Autonomous production scheduling |
Product delivery |
1) Dynamic updating of scheduled delivery milestone nodes based on production line resource allocation; 2) Automatic dynamic scheduling and task assignment based on production disturbances and real-time production progress |
Quality supervision |
Not involved |
Production safety |
Not involved |
Production line operation and maintenance visualization |
Product delivery |
1) Visualization of production schedule progress and delivery risks; 2) Visualization of personnel matching adequacy and equipment efficiency data; 3) Visualization of bottleneck resource and bottleneck process statistics; 4) Visualization of key material supply plan and shortage warning information |
Quality supervision |
1) Visualization of quality event statistics; 2) Visualization of personnel post competency index; 3) Visualization of overdue data for key production materials, equipment, and tooling fixtures |
Production safety |
1) Visualization of personnel and equipment load management data; 2) Real-time monitoring and warning data visualization of explosives temporary storage areas and explosive storage room equivalents; 3) Visualization of abnormal data on environmental temperature and humidity, key equipment operating parameters, and energy consumption |
Dynamic production line improvement |
Product delivery |
1) Iterative optimization of production line resource allocation based on bottleneck resource analysis data; 2) Dynamic optimization of process layout, warehouse layout, and logistics routes based on simulation analysis data |
Quality supervision |
1) Dynamic improvement of process flow and control parameters based on product quality analysis data; 2) Dynamic improvement of inspection and testing methods for high-quality-risk production links |
Production safety |
1) Dynamic improvement of technical safety control measures for production links with high safety risks |
3.1. Information Architecture Model
A digital twin is jointly constituted by the physical entity and the information model, wherein the information model serves as the fundamental basis of the digital twin. The precise mapping of the information model and the rapid control feedback of the physical entity constitute the core content of the digital twin. The intelligent transformation of the igniter production line based on digital twin is required to support multiple functional attributes, including data synchronization, data analysis, data-driven operation, virtual-physical mapping, and control decision-making. According to the digital twin application requirements of the igniter production line, the compositional characteristics of the existing information systems and physical entities of the igniter production line were analyzed, and a production-line-level digital twin information model comprising the virtual model layer, data assurance layer, analysis and computation layer, and system application layer was constructed (as shown in Figure 1), providing programmatic guidance for the application of digital twin technology to the igniter and analogous production lines. The business characteristics of each layer of this information model are described as follows:
1) Virtual Model Layer
This layer is employed to construct virtual models of all production areas, personnel, equipment, materials, and logistics within the igniter production line, as well as the associated business logic models (encompassing production plan propulsion, process route execution, in-process quality control, in-process safety control, in-process logistics control, and production material kitting). It achieves virtual integrated mapping of the physical entities of the igniter production line and their production control activities, and returns iterative data of symbiotic collaboration with physical entities to the data-driven layer.
2) Data-Driven Layer
Serving as the driver and engine of the digital twin model, the data-driven layer is responsible for perceiving and organizing activity, event, and status data from various production factors of the physical production line; processing iterative data fed back from the virtual model layer; providing driving data for the analysis and computation layer and managing feature data processed by the analysis and computation layer; acquiring configuration data from business modules of the application terminal layer; and aggregating business data from information systems including MES, ERP, SCADA, WMS, and the data operation and maintenance platform. Thereby, it provides an information framework for the integration and application of multi-source heterogeneous data of the igniter production line.
3) Analysis and Computation Layer
The analysis and computation layer constitutes the core layer for the functional realization of digital twin technology. It integrates algorithms required for functional kernels such as data mining, data analysis, and scheduling, processes behavioral decision information from the application terminal layer, and provides data service support for business modules of the application terminal layer according to specific decision contents.
4) Application Terminal Layer
From three aspects—visualization, simulation analysis, and production control—this layer provides business modules for production process control, dynamic planning and scheduling, process path planning, equipment health management, product quality assurance, and energy efficiency optimization analysis of the production line. It dispatches configuration data of various business modules to the data-driven layer, and transmits execution information to physical entities according to user instructions to achieve comprehensive control of all production line factors.
Figure 1. Information architecture model of the digital twin-enabled igniter production line.
3.2. Production Factor Model
Real-time production process data, including quality data, working-hour data, progress data, inspection data, and completion data, together with real-time perception data, are associated with and mapped to the production factor model of the production line digital twin. This mapping enables the digital twin to accurately reflect the actual operating status and progress of manufacturing resources, products, production plans, and process flows in the physical workshop. As a result, virtual-to-physical reflection and dynamic visual monitoring of all production factors and the entire production process can be achieved.
The effectiveness of digital twin applications is strongly related to the richness, timeliness, and completeness of full-factor production data. Therefore, it is necessary to construct production factor models that ensure unique identification of production factors, including man, machine, material, method, and environment, commonly referred to as 4M1E. These models provide explicit data tags for data acquisition, data governance, and data analysis of the igniter production line. The entity objects, digital encodings, and digital carriers involved in each production factor are listed in Table 2.
Table 2. Production factor model for the igniter production line.
Production Element |
Physical Object |
Digital Code |
Digital Carrier |
Human |
Operator; inspector; administrator; dispatcher; external personnel |
Employee ID |
RFID tag/card; face recognition; fingerprint |
Machine |
Automated equipment; explosion-proof box;
propellant AGV; non-propellant AGV |
Equipment ID |
QR code label; RFID tag/card |
Material |
Components; raw materials; propellant;
charge assembly; mold; tooling/fixture |
Material ID |
Barcode; QR code; RFID tag/card |
Method |
Process label; structured process |
Process code |
Barcode; QR code |
Environment |
Process layout; temperature and humidity sensor;
energy consumption sensor; vision sensor |
Area code |
QR code; RFID tag/card; MAC address |
3.3. Process Layout Model
Based on the structural characteristics and process flows of the production objects, the layout configurations of manufacturing cells encompass flow-line, cellular, fixed-position, and pulsating layouts, among others. The cellular layout organizes identical machine equipment and production functions within the same production operation unit according to the product process flow, i.e., constructing operation areas and arranging layouts based on the characteristics of completing similar activities or functions. This configuration exhibits high equipment utilization, strong equipment versatility, high flexibility in personnel allocation, and strong adaptability to process route variations.
Owing to the compact size characteristics of igniters, and because critical processes such as powder weighing, powder loading and pressing, and electrical performance testing align with the attributes of cellular layout, while also considering the operational continuity between processes and the connectivity of material flow, the process layout principle for the igniter production line was established as cellular layout as the primary configuration, supplemented by flow-line layout.
The igniter production line is partitioned into production units according to the product process flow. Based on the functional roles of workshop rooms, the workshop is divided into propellant warehouse, propellant temporary storage room, semi-finished product warehouse, finished product warehouse, powder weighing room, powder loading and pressing room, crimping room, electrical performance testing room, and non-explosive assembly room, among others. Concurrently, a flow-line layout is implemented according to the complete product process flow. On the basis of automated material handling, the functional zones of the production line are interconnected through automated material handling, with connection points established at each functional unit to form linkages among the component preparation area, production area, and warehouses of the production line. Subsequently, the overall logistics route planning is conducted based on the process layout.
3.4. Manufacturing Process Model
The igniter production line relies on real-time production planning and resource scheduling data to drive production and logistics activities. The definition of structured process data with clear hierarchy, rational structure, and explicit relationships is critical for establishing the mutual mapping between the virtual and physical production lines and successfully implementing the digital twin of the igniter production line. It also serves as the foundation for autonomous planning, scheduling, and resource allocation. Therefore, it is necessary to construct product-structured process data that conforms to process requirements, reflects the sequential relationships among process contents, characterizes the actual production takt of products, binds the resource configuration requirements and constraints of each process, and encompasses key production factors. The emphasis lies in quantifying general parameters such as product process operation sequence, progress offset relationships between processes, human-machine operation efficiency of each process, and material and resource demand quantities of each process, as well as personalized parameters reflecting production organization modes such as curing time, production preparation time, process outsourcing time, and continuous operation. Furthermore, resource data required for each process must be bound. The key field definitions and quantitative information required for the product structured process data, as specified according to the aforementioned requirements, are presented in Table 3.
Table 3. Manufacturing process model for structured process data.
Field |
Quantitative
Range |
Step |
Unit |
Characteristic Attribute |
Operation sequence number |
1 - ∞ |
1 |
— |
Indicates the operation order of the process |
Operation type |
0 - 1 |
1 |
— |
Indicates the type of operation (1 = outsourced operation) |
Offset progress |
0% - 100% |
0.01 |
— |
Defines the progress offset between the current operation and the previous operation |
Production unit |
— |
— |
— |
Binds the current operation to its corresponding production unit |
Work center |
— |
— |
— |
Binds the current operation to its corresponding work center |
Unit manpower |
1 - ∞ |
1 |
— |
Binds the unit manpower resources required by the current operation |
Maximum manpower |
1 - ∞ |
1 |
— |
Binds the maximum number of manpower resources allowed for the current operation |
Manual operation efficiency |
1 - ∞ |
1 |
s |
Defines the manual operation efficiency value of the current operation |
Equipment operation efficiency |
1 - ∞ |
1 |
s |
Defines the equipment operation efficiency value of the current operation |
Material code |
— |
— |
— |
Binds the material required by the current operation |
Material quantity |
1 - ∞ |
1 |
— |
Defines the quantity of materials required by the current operation |
Tooling/fixture code |
— |
— |
— |
Binds the tooling/fixture required by the current operation |
Tooling/fixture quantity |
1 - ∞ |
1 |
— |
Defines the quantity of tooling/fixtures required by the current operation |
Solidification time |
0 - ∞ |
1 |
s |
Defines the idle/curing time after completion of the current operation |
Production
preparation time |
0 - ∞ |
1 |
s |
Defines the preparation time before the current operation starts |
Outsourcing time |
0 - ∞ |
1 |
h |
Defines the estimated completion time of the current outsourced operation |
Continuous operation |
0 - 1 |
1 |
— |
Indicates the continuity relationship between the current operation and the preceding/following operations |
3.5. Data Application Model
The data application model serves as the business carrier for achieving the targeted enhancements of digital twin technology application in the igniter production line. It primarily performs risk analysis and assessment of production line risks—including personnel and equipment anomalies, production task delays, equipment maintenance failures, and production resource shortages—based on real-time interactive virtual-physical mapping data. This enables current production control, future simulation-based prediction, and intuitive retrospective tracing of past events, thereby providing information services for users at different hierarchical levels and diverse business application requirements.
From the perspective of targeted enhancements enabled by digital twin technology application, the key factors of the igniter production line requiring dynamic monitoring primarily encompass operators, critical equipment, operation areas, inventory zones, and production line logistics. Risk analysis is conducted from three dimensions: product delivery, quality, and safety. The specific contents of dynamic monitoring and risk analysis to be achieved for each key factor are presented in Table 4.
4. Implementation
Based on the aforementioned ideal information model for the igniter production line, the implementation of production-line-level digital twin primarily encompasses the following tasks: constructing a comprehensive factor perception and data service middleware that embodies the production factor model; developing autonomous planning and resource scheduling algorithms interfaced with the manufacturing process model; establishing a dynamic monitoring and risk analysis platform that carries the data application model; and implementing a digital twin interaction platform that realizes the information architecture model.
The implementation scope covered one digitalized igniter assembly production line, including non-explosive assembly, powder weighing, powder loading and pressing, crimping, electrical performance testing, material storage, work-in-process transfer, and key area safety monitoring. The pilot deployment covered one complete production cycle, spanning production-plan release, process execution, quality inspection, logistics transfer, and completion feedback. The operational KPIs used to judge implementation success included data acquisition completeness, data synchronization latency, schedule-feasibility rate, abnormal-event response time, warning accuracy, manual rescheduling workload, and traceability completeness for personnel, equipment, material, method, and environment records.
Table 4. Data application model for dynamic monitoring and risk analysis.
Key Element |
Dynamic Monitoring |
Risk Analysis Content |
Operator |
Personnel protective status |
Safety risk |
Location within the area |
Safety risk |
Operation execution status |
Quality risk |
Post competency status |
Quality risk |
Post suitability condition |
Quality risk |
Production task |
Production task progress |
Product delivery risk |
Quality consistency |
Quality risk |
Key equipment |
Equipment protective status |
Safety risk |
Equipment operating status |
Safety risk/quality risk |
Equipment control parameter status |
Quality risk |
Equipment effective status |
Quality risk |
Operation execution status |
Product delivery risk |
Work area |
Production schedule progress |
Product delivery risk |
Environmental temperature and humidity status |
Quality risk |
Personnel count |
Safety risk |
Grounding facility status |
Safety risk |
Equivalent quantity status |
Safety risk |
Inventory area |
Material validity status |
Quality risk |
Inventory equivalent quantity status |
Safety risk |
Material stocking status |
Product delivery risk |
Storage capacity status |
Product delivery risk |
Production line logistics |
Load-carrying vehicle status |
Safety risk |
Propellant-carrying vehicle status |
Safety risk |
Material flow load status |
Product delivery risk |
4.1. Comprehensive Factor Perception and Data Service
Middleware
4.1.1. Comprehensive Factor Perception of Production Line
The comprehensive factor perception of the igniter production line is primarily implemented through adaptive upgrades of existing information systems, including SCADA, WMS, temperature and humidity monitoring systems, equipment host computers, visual inspection platforms, and energy consumption monitoring terminals. Leveraging hardware facilities such as fingerprint modules, RFID readers and writers, facial recognition modules, data acquisition gauges and tools, workstation terminals, barcode scanners, keyboards, cameras, industrial cameras, PLCs, and various types of sensors, data interaction and acquisition are conducted in conjunction with communication protocols including TCP/IP and RS-485 (all employing relational databases for structured data storage).
4.1.2. Data Service Middleware
The data service middleware serves as the information carrier of the data-driven layer within the production-line-level digital twin information architecture model. It is responsible for the aggregation, association, integration, and cleansing of multi-source heterogeneous data from all production line factors, as well as the integration of real-time acquired data, planning and scheduling data, and autonomous decision-making data, thereby providing data-driven services for the digital twin interaction platform. It interfaces with underlying production line information systems such as SCADA, WMS, temperature and humidity monitoring systems, equipment host computers, visual inspection platforms, and energy consumption monitoring terminals, as well as upper-level information systems including APS and MES. The acquired data from each information system are stored in real time in the corresponding databases. The data service middleware performs data governance, integration, and push according to the perception of business requirements of each production factor. The overall business architecture is illustrated in Figure 2.
The data pipeline was implemented as a unified object-event-record flow. The main data sources included SCADA equipment status and process parameters, WMS inventory and equivalence records, MES/production-plan orders, temperature and humidity monitoring data, equipment host-computer logs, visual inspection results, energy-consumption terminals, RFID/barcode/fingerprint/face-recognition devices, AGV logistics records, and manual confirmation records from workstation terminals. Update modes were configured according to data criticality: equipment status, safety-area access, environmental alarms, and AGV location were acquired in real time; inventory balance, tooling status, and energy data were synchronized periodically; production-plan release, quality inspection completion, order insertion, process suspension, and abnormal-event confirmation were handled as event-driven records.
Each object record was bound to a unique identifier, including employee ID, equipment ID, material ID, process code, area code, batch number, work-order number, tooling/fixture ID, inspection record ID, and logistics task ID. Missing records were handled through source-side retry, middleware-level null-state labeling, and manual review when a safety- or quality-related field was absent. Conflicting records were resolved according to a predefined priority order: safety interlock and equipment host-computer records had the highest priority, followed by inspection records, WMS/MES records, perception-device records, and manual confirmation records. All corrections retained the original timestamp, source identifier, correction time, and operator identity to support traceability.
Figure 2. Business architecture of data service middleware.
4.2. Planning and Scheduling Algorithm Model
Enhancing the scheduling planning and management capabilities of the igniter production line for production tasks, material resources, equipment resources, and human resources through digital twin as the technical means constitutes the key objective of this study. The overarching requirement for implementing autonomous planning and resource scheduling in the igniter production line is as follows: based on production line planning, driven by dynamic perturbation events, and comprehensively considering factors including task priority, workstation capability, balanced production, resource kitting, as well as overlapping, concurrent, and parallel operations in the production process, dynamically generating operation plans specific to process steps that conform to the actual production organization mode of the production line. This enables refined scheduling of the production line, as well as delivery cycle assessment based on scheduling results, bottleneck resource identification and optimization, human-machine load adjustment and control, material shortage early warning, and material inventory control. Therefore, the algorithm is required to solve locally optimal solutions conforming to current actual production constraints within multiple constraint rules according to pre-established multiple optimization objectives, relying on high-efficiency algorithmic models [15]-[17].
The scheduling algorithm model is primarily composed of optimization algorithms and rule strategies. The optimization algorithms are mainly employed to enhance the convergence speed and accuracy of the scheduling algorithm model, obtaining scheduling results that satisfy optimization objectives. The rule strategies are utilized to ensure that scheduling results conform to the actual production organization mode of the production line and meet quality and safety control requirements of the production process. Whether the output results of the scheduling algorithm model conform to the production organization mode of the production line directly determines the success or failure of intelligent scheduling technology application. Executable scheduling results also serve as the prerequisite for conducting bottleneck-based production line resource configuration optimization. Therefore, the algorithmic model focuses on constructing rule strategies that conform to the manufacturing mode of the igniter production line.
Regarding optimization algorithms. Due to the particularity of the manufacturing domain, the aerospace manufacturing industry prioritizes product delivery as the primary objective, with secondary objectives such as balanced production, load management, and cost control considered on the basis of ensuring product delivery. Given the complexity and diversity of production line planning and scheduling problems, as well as the problem adaptability of different algorithms, the optimization algorithm integrates optimal algorithms, including linear programming and dynamic programming, and heuristic algorithms, including genetic algorithms and simulated annealing algorithms, according to the complexity of the production line scheduling problem mapped from optimization objectives and constraint conditions. Linear programming and dynamic programming ensure the computational efficiency of the algorithmic model, while genetic algorithms and simulated annealing algorithms concurrently account for the global search capability and global convergence of the algorithmic model.
1) Regarding rule strategies. Based on the characteristics of igniter products, the following rule strategies have been established in the scheduling process to conform to the production organization mode and quality control measures:
Small-batch overtime: If the remaining man-hours of the current production plan exceed the working hours of the current day but do not exceed a preset duration, overtime is initiated on the current day to ensure product quality consistency and production continuity within the same batch.
Small-batch changeover: Within a specific duration before the end of the working period, the current production line prioritizes scheduling processes that can be completed on the current day and whose subsequent process is curing or outsourcing, thereby optimizing the production takt.
Continuous operation: For multiple consecutive processes with sequential process index numbers configured for continuous operation, they are treated as a single process for scheduling purposes to ensure production continuity across multiple processes.
Process alignment: For continuous operation processes with configured process alignment, the completion time of each process must be ensured to be consistent to conform to the actual production mode.
Resource over-allocation: By setting additional equipment and personnel quantities for work centers, extra resources are permitted to be activated when scheduling results fail to meet delivery deadlines, providing a reference for resource over-allocation of the production line.
Intra-batch resource occupation retention: The same batch of products is constrained to complete related processes using identical equipment and human resources, with prohibition of replacement or adjustment midway, to ensure product quality stability and consistency.
Preparation time recalculation across days: For processes with configured preparation time, if the operation spans multiple days, the preparation time must be recalculated to conform to the actual production mode and enhance the accuracy of scheduling results.
Co-production restriction for identical products: The same process of different batches of identical products is restricted from being produced on the same day to prevent batch mixing.
Based on the aforementioned content, the computational mode adopted by the algorithmic model (as illustrated in Figure 3) can be summarized as follows: with product delivery as the primary objective, considering the priority of various secondary optimization objectives, and according to the selected operational rule strategies, an iterative verification computation mode with forward-backward associated constraints is performed under the restrictions of various constraint information. Based on the prerequisite constraints required to achieve product delivery, sequential forward computation is conducted. When intermediate verification reveals that forward-backward constraints cannot be satisfied, the computation returns to the node where the delivery deadline is not met for plan adjustment, after which forward computation continues. This process is repeated iteratively until all constraints and delivery objectives are satisfied, forming an optimal production plan that satisfies scheduling rules under current constraint limitations. During this process, if the delivery objective cannot be fully satisfied after iterative verification, capacity load analysis results are output, and corresponding resolution strategies are provided to furnish quantitative reference data for production line resource allocation and plan adjustment.
![]()
Figure 3. Planning and scheduling algorithm model.
4.3. Digital Twin Interactive Platform
The digital twin interaction platform serves as the core technical platform embodying the principle of “physical-to-virtual reflection and virtual-to-physical control”, providing intuitive visual information services for real-time control and dynamic optimization of production line activities. To achieve bidirectional mapping and real-time interaction between virtual and physical models across the production line, 3D model libraries are constructed for models with physical entity characteristics—including production lines, personnel, equipment, storage fixtures, and transport carriers—for invocation by the dynamic 3D visualized interaction module of the application terminal layer. For business logic models that cannot be physically visualized, such as production plan propulsion, process route execution, in-process quality control, in-process safety control, in-process logistics control, and production material kitting, visualized interaction modules based on operational data are employed for display and early warning, grounded in key characteristic datasets reflecting the corresponding business logic. The implementation primarily encompasses the operational data visualization module and the 3D visualized interaction module.
The closed-loop control path distinguishes three output types. Advisory outputs include schedule-risk prompts, load-balance suggestions, equipment-maintenance recommendations, and quality-traceability reminders, which are displayed to dispatchers or process engineers for reference. Approval-required outputs include production-task resequencing, overtime activation, cross-day preparation-time recalculation, resource over-allocation, and release of abnormal-quality holds, which must be confirmed by authorized personnel before execution. Automatically executable outputs are limited to bounded actions such as data refresh, warning push, dashboard update, task-status synchronization, visualization linkage, and execution of predefined safety interlock notifications.
The operational prerequisites for transferring this framework to other production lines are stable object coding, accessible equipment or workstation data interfaces, clear process routes and qualification rules, validated safety and quality thresholds, and an authorization mechanism for human approval. The boundary condition is that the framework supports decision assistance and bounded execution within predefined process and safety rules; it does not replace certified safety interlocks, quality acceptance decisions, or manual approval where regulations require human responsibility.
Regarding operational data visualization. Targeting the business logic models required for virtual-physical mapping of the production line, and building upon the foundation of dynamic monitoring and risk analysis of key factors, early warning trigger mechanism codes for various risk analysis operations are developed based on integrated data from the data service middleware. Consequently, the operational data visualized interaction module for the igniter production line is constructed, with its interface illustrated in Figure 4.
Figure 4. Operational data visualization module.
The risk-analysis logic adopts a threshold-and-score scheme. Delivery risk is triggered by milestone deviation, excessive remaining workload, unavailable equipment/personnel, or material-kitting shortage. Quality risk is triggered by missing inspection records, mismatch between operator qualification and process requirements, deviation of key process parameters, inconsistent tooling/material binding, or abnormal inspection results. Safety risk is triggered by unauthorized personnel entry, missing protective-status confirmation, hazardous-area equivalence excess, environmental parameter overrun, equipment alarm, or abnormal logistics access.
Alert severity is assigned by combining event criticality, deviation magnitude, duration, and propagation range. A low-level warning indicates a recoverable deviation that does not affect the current milestone or safety boundary; a medium-level warning indicates that manual confirmation or local rescheduling is required; a high-level warning indicates violation or imminent violation of safety, quality, resource, or delivery constraints and triggers escalation to authorized personnel. For safety-critical events, the highest severity is assigned directly regardless of the aggregate score.
Figure 5. 3D visualized interaction module.
Regarding dynamic 3D visualization of the production line. Targeting the twin models required for virtual-physical mapping of the production line, 3D models of critical equipment and workshop layouts are constructed based on 3D configuration development tools, with model operation animation configurations and inter-model event linkage configurations established. In this way, the virtual scene of the digital twin production line is constructed. Real-time data provided by the data service middleware are refreshed through HTTP polling, and historical playback of the digital twin scene is realized. The production line 3D visualized interaction module is illustrated in Figure 5.
5. Validation Results
The pilot validation focused on whether the proposed framework could be deployed on the target igniter production line and whether it could support stable production control under real workshop constraints. The validation scope included perception access for key 4M1E objects, middleware integration of multi-source heterogeneous data, scheduling computation under released production plans and disturbance events, risk-warning generation for delivery/quality/safety events, and visual interaction through the operational data and 3D visualization modules.
The baseline mode was manual production-plan adjustment supported by independent information systems and offline communication among dispatching, quality, warehouse, and workshop personnel. The digital-twin pilot was evaluated using the same operational KPI categories defined in the implementation section: data acquisition completeness, synchronization latency, schedule-feasibility rate, abnormal-event response time, warning accuracy, manual rescheduling workload, and traceability completeness. The validation confirmed that the framework could connect the required production objects, generate feasible schedules under hard safety and quality constraints, expose delivery/quality/safety risks in the interactive platform, and provide traceable closed-loop records from event occurrence to human approval or bounded automatic execution.
Quantitative KPI records further supported the pilot validation. The completeness of automatically acquired 4M1E records reached more than 95%. Key equipment and environmental data were synchronized within 0.5 s. For typical disturbance events, manual rescheduling time was reduced from approximately 30 min to 2 min. Traceability records covered personnel, equipment, material, process, and environmental objects for all pilot work orders.
These results support the implementability of the proposed technical route at the production-line level. Because the pilot was conducted on one igniter assembly line and was constrained by the available interfaces of existing equipment and information systems, transferability should be understood as architectural and methodological transferability rather than direct plug-and-play replication. Additional multi-line deployment and longer-period quantitative comparison are still required to evaluate statistically robust performance improvement.
6. Conclusion
The study takes all production factors of the igniter production line as the research object and digital twin as the technical approach, conducting work encompassing the clarification of construction objectives for the igniter production line, the design of the ideal information model for the production line, implementation of the production line digital twin, and pilot validation of the proposed technical route. The relatively generalized concept of digital twin in manufacturing processes has been concretized at the production-line level, yielding a series of technical achievements, including the production-line-level digital twin information architecture model, product structured process dataset, comprehensive factor perception methodology and data service middleware, autonomous planning and resource scheduling model, dynamic monitoring and risk analysis methodology, and production line digital twin platform. The validation results indicate that the proposed framework is implementable under the available data interface, process rule, and human approval conditions of the target igniter production line. Its transferability is mainly reflected in the reusable architecture, object coding logic, scheduling-constraint formulation, risk-warning logic, and closed-loop control path, while line-specific thresholds, equipment interfaces, process rules, and safety authorization mechanisms must be reconfigured before application to other discrete manufacturing lines.