Furniture factory production schedule optimization method based on digital twinning

Abstract

A furniture factory production scheduling optimization method based on digital twinning integrates multi-dimensional data of machines, materials, environment and personnel, covers four factors of man-machine object ring of a factory, provides a complete data basis for production scheduling, realizes fine management of furniture factories, and enables the production process to be intelligent, informationized, processized and normalized; the synchronization of the physical entity and the virtual model is realized through the digital twin model, the virtual model and the physical entity are mapped, and the state of the physical entity is reflected in real time; real-time sensing data is learned based on a neural network model and a machine learning algorithm, so that a production scheduling scheme better meets actual production requirements; the production scheduling is dynamically optimized based on real-time sensing data, emergency situations such as machine faults, environment abnormity and resource configuration abnormity can be quickly responded, production delay is reduced, and virtual-real interaction, data driving and intelligent decision making of a factory are achieved through the digital twinning technology.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of intelligent manufacturing and digital twinning, and particularly relates to a furniture factory production scheduling optimization method based on digital twinning. BACKGROUND

[0002] In modern furniture factory production management, the rationality of production scheduling directly affects production efficiency, resource utilization, safety risk, and production cost, etc. Traditional production scheduling methods rely heavily on human experience, which has the following defects:

[0003] 1. Incomplete data perception, lack of systematic integration: Existing methods focus on single-dimensional data such as machine status or material inventory, ignoring factors such as environmental temperature and humidity, dust concentration, etc. that affect equipment precision, material performance, and personnel efficiency. The dynamic matching of personnel skill level, real-time working state and scheduling also lacks data support, resulting in a disconnect between the scheduling and the actual complex and variable production scene.

[0004] 2. Model and physical world are disconnected, prediction accuracy is insufficient: Traditional scheduling models are mostly static models that cannot form real-time and dynamic mapping relationships with physical factory entities. When equipment performance declines, material supply fluctuates, or environmental abnormalities occur, it cannot be perceived and corrected in a timely manner, resulting in a deviation of prediction results based on historical data or theoretical assumptions from reality, which can easily cause production delays.

[0005] 3. Dynamic response and adjustment capability lag: In the face of equipment failure, material shortage or environmental parameter exceeding, etc. disturbance events, existing systems usually rely on manual intervention for scheduling adjustment. This process relies on individual experience and lacks global optimization analysis under data-driven conditions, making it difficult to quickly generate a scientific and feasible adjustment plan, which can easily cause production delays and resource waste.

[0006] 4. Decision-making loop is not formed, intelligent level is limited: In existing solutions, the perception, analysis, decision-making, and execution steps often operate in isolation and do not form an automatic feedback optimization loop. After decision-making, there is a lack of real-time feedback to verify and optimize the model, and the intelligent level needs to be improved. There is still a gap between the goal of integrated intelligent production of "perception-decision-execution" and the current situation. SUMMARY

[0007] The purpose of the present application is to propose a furniture factory production scheduling optimization method based on digital twinning, which combines digital twinning technology with furniture factory production and manufacturing, perceives data from four dimensions, integrates real-time data from each monitoring point, optimizes the prediction of each monitoring point and production scheduling based on the digital twinning model, and generates corresponding decisions to achieve dynamic adjustment and effective optimization of production scheduling plans. The method solves the problems of incomplete data perception and weak dynamic adjustment capability in existing production scheduling methods, and improves the efficiency and progress of the production process.

[0008] To achieve the above functions, the application designs a furniture factory production scheduling optimization system based on digital twinning, including a physical entity layer and a digital twinning layer;

[0009] The physical entity layer includes a physical perception module and a production control module; the physical perception module includes a perception object unit and a perception device unit; the production control module includes a decision receiving unit;

[0010] The digital twinning layer includes a digital twinning module; the digital twinning module includes a DT modeling unit, a DT updating unit, a prediction and optimization unit, and a decision making unit;

[0011] The perception object unit is each monitoring point in the furniture factory environment, the perception device unit is used for real-time state monitoring of each point in the physical scene of the furniture factory, collecting relevant data of the perception object unit, performing multi-source data perception, data preprocessing and standardization, and uploading effective data to the DT modeling unit of the digital twinning layer; the DT modeling unit receives data and integrates to obtain multi-dimensional perception data, performs furniture factory modeling and data correlation, and establishes a real-time mapping relationship between data and models, i.e. a digital twinning model; the DT updating unit updates and corrects the parameters of the digital twinning model in time for maintenance; the prediction and optimization unit performs production condition prediction and optimization analysis of the furniture factory based on multi-dimensional perception data; the decision making unit formulates furniture factory production scheduling optimization decisions according to the prediction and optimization analysis results; the decision receiving unit receives decisions issued by the decision making unit of the digital twinning layer and sends them to the perception object unit for execution.

[0012] The furniture factory production scheduling optimization system method based on the above system includes the following steps:

[0013] Step S1: Based on the perception data of four monitoring points of production machines, production materials, workshop environment, and production personnel, the running state data of factory equipment, material out-of-stock, in-stock and inventory data, environmental parameters, and personnel scheduling and work content data are collected by the perception device unit of the physical entity layer, the collected raw data are preliminarily processed, including data preprocessing and standardization, and the processed real-time state information is uploaded to the DT modeling unit of the digital twinning layer;

[0014] Step S2: The DT modeling unit of the digital twinning layer receives and stores the real-time state information processed by the perception device unit, integrates multi-dimensional data from different perception devices, extracts and selects features from the integrated data, constructs digital twinning models of machines, materials, environment, personnel, and production scheduling in the DT modeling unit respectively, realizes real-time mapping of physical entities and virtual models, and updates the digital twinning model in the DT updating unit to keep the digital twinning model and the physical entity state real-time synchronized;

[0015] Step S3: In the prediction optimization unit, based on the neural network model and the machine learning algorithm, the maintenance requirements are predicted according to the equipment operation data, and the maintenance warning is generated; the material remaining situation is predicted according to the in-out warehouse situation, and the shortage risk is warned; the influence of the environmental parameters on the machine and the material is predicted according to the real-time temperature and humidity, and the warning is given when the threshold is exceeded; the personnel state is recorded to provide data support for production efficiency prediction; the production scheduling scheme is predicted according to the real-time data of the four monitoring points; the prediction optimization unit sends the generated maintenance warning, material inventory warning, environmental control warning, production efficiency prediction and production scheduling scheme prediction to the decision-making unit;

[0016] Step S4: In the decision-making unit, decisions related to machine equipment, material inventory, environmental temperature and humidity, personnel arrangement and production scheduling scheme are generated; the decisions generated by the prediction optimization unit are combined to dynamically adjust the parameters of each monitoring point and the production scheduling scheme, and the decisions are sent to the decision receiving unit of the physical entity layer;

[0017] Step S5: The sensing object unit of the physical sensing module of the physical entity layer executes the decisions from the decision receiving unit, and feeds back the execution results to the sensing device unit, which uploads the feedback data to the digital twin layer for subsequent model updating and optimization, forming a closed-loop process of sensing-modeling-prediction-decision-making-execution-feedback.

[0018] Compared with the prior art, the above technical scheme has the following technical effects:

[0019] (1) Comprehensive data sensing: Integrating machine, material, environment and personnel multi-dimensional data, covering the four elements of "man-machine-material-environment" in the factory, providing complete data basis for production scheduling, realizing fine management of furniture factory, and making production process intelligent, information-based, process-based and standardized.

[0020] (2) High prediction accuracy: The digital twin model realizes the synchronization of physical entity and virtual model, the virtual model is mapped with the physical entity, and the state of the physical entity is reflected in real time; based on the neural network model and the machine learning algorithm, the learning of real-time sensing data makes the production scheduling scheme more suitable for actual production needs.

[0021] (3) Strong dynamic adjustment capability: Based on real-time sensing data, the production scheduling is dynamically optimized, which can quickly respond to machine failure, environmental abnormality, resource allocation abnormality and other sudden situations, and reduce production delay.

[0022] (4) High degree of intelligence: Integrating real-time data of monitoring points such as machine, material, environment and personnel, combining algorithm model, generating decisions based on prediction optimization of each monitoring point and production scheduling, dynamically adjusting production scheduling scheme, realizing virtual-real interaction, data-driven and intelligent decision-making of factory through digital twin technology. Attached Figure Description

[0023] Figure 1 This is an overall architecture diagram of a specific embodiment of the present invention.

[0024] Figure 2 This is a diagram of the physical entity layer architecture in a specific embodiment of the present invention.

[0025] Figure 3 This is a diagram of the digital twin layer architecture in a specific embodiment of the present invention.

[0026] Figure 4 This is a flowchart of the optimized production scheduling operation of a furniture factory based on digital twins, as described in a specific embodiment of the present invention. Detailed Implementation

[0027] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings.

[0028] like Figure 1 As shown in the figure, this embodiment presents a system architecture for a furniture factory production scheduling optimization method based on digital twins. This system architecture can be divided into two parts: a physical entity layer and a digital twin layer. The physical entity layer consists of a physical sensing module and a data production control module, while the digital twin layer consists of digital twin modules.

[0029] The physical perception module is responsible for real-time status monitoring of key points in the physical scene of the furniture factory, collecting multi-source data, preprocessing and standardizing data, executing decisions sent by the decision-making unit, and feeding back the processed real-time status information and execution results to the DT modeling unit of the digital twin layer.

[0030] The production control module is responsible for sending the decisions transmitted by the decision-making unit in the digital twin layer to the sensing object unit of the physical sensing module. The sensing object unit executes the decisions to control the physical production process.

[0031] The digital twin module is responsible for receiving the processed real-time state information sent by the physical entity layer, integrating multi-dimensional data from different sensing devices, extracting and selecting the fused data, associating with the geometric model mapping, constructing the corresponding digital twin, and updating and correcting the digital twin model parameters in time by comparing the model prediction value with the actual sensing data, maintaining the real-time mapping relationship between the physical entity and the virtual model, and keeping the digital twin model and the physical entity in real-time synchronization; using the neural network model, the future state of the equipment failure rate, the material inventory, the environmental temperature and humidity, and the personnel arrangement is predicted, and the production process is optimized and analyzed in combination with the actual production target, such as machine equipment parameter optimization, workshop environment parameter optimization, production material inventory optimization, personnel arrangement optimization, and production scheduling optimization; using the machine learning algorithm to predict the production scheduling, combining the real-time data and the prediction optimization, dynamically adjusting the decision of designing each monitoring point and the production scheduling, and sending the above decisions to the production control module of the physical sensing layer.

[0032] As shown in Figure 2 The physical entity layer of the present application is composed of a physical sensing module and a production control module. The physical sensing module is responsible for real-time state monitoring of each point of the furniture factory physical scene, multi-source data acquisition, data preprocessing and standardization, execution of decisions issued by the decision-making unit, and feedback of the sensed real-time state information and execution results to the DT modeling unit of the digital twin layer; the production control module is used to send the decisions delivered by the decision-making unit in the digital twin layer to the sensing object unit.

[0033] As shown in Figure 3As shown, the digital twin layer of the present application comprises a digital twin module, wherein the digital twin module comprises a DT modeling unit, a DT updating unit, a prediction optimization unit and a decision making unit. The DT modeling unit is used to receive and store the pre-processed and standardized real-time data uploaded from the perception device unit in the physical entity layer, integrate multi-dimensional data from different perception devices, perform feature extraction and selection on the fused data, and provide reliable and accurate data source for model construction. A geometric model of the furniture factory is constructed by a three-dimensional modeling software, the processed real-time state information is associated with the geometric model, and a real-time mapping relationship between the physical entity and the virtual model is established, thereby providing model support for subsequent updating, prediction optimization and decision making; the DT updating unit is used to receive multi-dimensional perception data processed by the DT modeling unit, update and correct the digital twin model parameters in a timely manner by comparing the model prediction value with the actual perception data, maintain the real-time mapping relationship between the physical entity and the virtual model, and keep the digital twin model and the physical entity in real-time synchronization; the prediction optimization unit is used to receive multi-dimensional perception data processed by the DT modeling unit, predict the future state of equipment failure rate, material inventory, environmental temperature and humidity, and personnel arrangement based on a neural network model, and at the same time, optimize and analyze the production process in combination with the actual production target, such as machine equipment parameter optimization, workshop environment parameter optimization, production material inventory optimization, personnel arrangement optimization and production scheduling optimization, and send the above prediction optimization to the decision making unit; the decision making unit is used to receive the prediction optimization transmitted by the prediction optimization unit, generate decisions related to equipment maintenance, parameter optimization, personnel configuration and production scheduling scheme of the monitoring points, and issue them to the decision receiving unit of the physical entity layer, and adjust the parameters of each monitoring point and the production scheduling scheme in real time based on the decisions, so as to guide the physical production process.

[0034] The present application introduces digital twin technology in the furniture factory production scene, constructs a furniture factory production scheduling optimization method based on digital twin according to steps 1 to 5, and realizes prediction of production scheduling under digital twin technology by using the method.

[0035] Step 1, based on the perception and control point data of machine, material, environment and personnel, the running state data of factory equipment, material out-of-warehouse, in-warehouse and inventory data, environmental parameters, and personnel scheduling, work content and other data are collected by the perception device unit of the physical entity layer, the collected raw data is preliminarily processed, including data preprocessing and standardization, and the real-time processed data is uploaded to the DT modeling unit of the digital twin layer.

[0036] In step 1, the man-machine-thing-environment framework is followed to associate people, machines, materials and environment with each other to improve the production efficiency of the furniture factory. In order to realize the informatization, process and standardization of the production process, the method needs to acquire machine running state data, material inventory data, temperature, humidity, air quality, and personnel scheduling time and work content in real time. To achieve this goal, the sensing end of the physical entity layer is deployed with vibration sensors, temperature sensors, speed sensors (for collecting equipment running state), RFID readers (for collecting material inventory), dust detectors (for measuring dust concentration), gas detectors (for measuring gas composition), and sensors related to environmental monitoring. For personnel data, through the interface of the factory manufacturing execution system (MES) or enterprise resource planning (ERP) system, the structured personnel scheduling table, skill profile, and post information are acquired in real time; at the same time, the on-site state and work progress of personnel can be collected through work terminal, card swiping / face recognition sign-in equipment, and behavior analysis module combined with camera, etc. These multi-source heterogeneous data are unified and gathered to the sensing device unit for preprocessing. For the above-mentioned multi-source heterogeneous data source categories, adapters or agents are used for access. For sensor data stream, access through industrial protocols such as MQTT, OPC UA; for system data, access through API interface, database connection, etc.; for PLC data, access through industrial gateway for protocol analysis and collection.

[0037] The sensed data are detected and processed from the physical environment and system or terminal, and all the data are completed within a fixed frame time τ to meet the real-time requirement. Different types of data need to be sensed by different sensors, and different sensors collect data at different sampling rates. The total cycle of the sensing device unit T is divided into T time slots, with a length of τ=T / N, and the time slot set is defined as n∈{1,2,......,N}. The data collected in the sensing process can be represented as:

[0038]

[0039] wherein η∈(0,1) represents the time allocation coefficient for dividing the data sensing and processing stages of the sensing device unit, τ represents the frame time, and k is defined as the K sensing devices of the sensing device unit, represents the sensing rate of the sensing device unit.

[0040] At the beginning of time slot n, the sensing device unit needs to collect D k [n] bits of sensing data for DT synchronization, D k [n]∈[D min ,D max ] bits, wherein D min and D maxare the minimum and maximum amount of sensing data required for DT synchronization, respectively. The perception data computation time can be expressed as:

[0041]

[0042] For the collection and processing of production personnel-related information, the standard data interface is connected with the manufacturing execution system, enterprise resource planning system, and workstation terminal to obtain structured data such as personnel basic archives, scheduling plans, on-duty states, workstation bindings, and production task execution progress in real time. After extraction, the data is encapsulated as standardized personnel state events with spatiotemporal tags in the edge server and uploaded to the digital twin layer through a unified message middleware and sensor data stream synchronization, and then the state, skill, and task attributes of the corresponding "personnel digital twin" in the virtual model are mapped and updated to provide real-time, structured data input for personnel arrangement optimization and production efficiency collaborative prediction.

[0043] In step 2, the DT modeling unit of the digital twin layer receives and stores real-time state information from the perception device unit processing. All perception data are time-stamped with high-precision timestamps by a unified time service before entering the DT modeling unit. Multi-dimensional data from different perception devices are integrated, and features are extracted and selected from the integrated data to construct new features more representative of prediction and optimization tasks from the original data, such as time-domain and frequency-domain features extracted from equipment vibration time series data); for environmental data, calculate the rolling average, change trend slope, etc. Feature selection uses tree model-based feature importance sorting to select the strongest feature subset related to the prediction target to reduce model complexity and prevent overfitting. In the DT modeling unit, digital twin models of machines, materials, environments, personnel, and production schedules are constructed, where physical entities refer to specific machine equipment, material inventory, workshop space area, production workers, and production orders being executed in the factory; virtual models are three-dimensional visual models and data-driven parameterized models established in the digital space, such as equipment performance degradation models, inventory change models, and production process models. The mapping relationship ensures that the state, attributes, and behavior of the virtual model can truly reflect and synchronize the real-time situation of its corresponding physical entity, realizing real-time mapping of physical entities and virtual models; the DT update unit updates the digital twin model in a timely manner to keep the digital twin model and the physical entity state real-time synchronized.

[0044] In step 2, the DT is deployed on the edge service, and the line-of-sight path from the physical entity to the edge server is blocked, so the channel from the physical entity to the edge server is modeled as a Rayleigh fading channel, which is represented as follows:

[0045]

[0046] where β0 is a path loss factor, p k with p B respectively represent the location of the sensor and the edge server, the Euclidean distance between sensor k and the edge server α represents the path loss exponent,

[0047] the achievable rate u at which the perception device transmits perception information to the edge server in time slot n k [n] can be expressed as:

[0048]

[0049] where B k =B / K is the bandwidth allocated to the perception device, B is the total bandwidth of the system, P k [n] is the transmit power of the perception device, σ 2 is the power of the Gaussian white noise.

[0050] The transmission delay of the uploaded sensing data corresponding to the synchronization task needs to be completed within a specified time length, which can be expressed as:

[0051]

[0052] The total DT delay time is composed of two parts, the data sensing delay and the transmission delay , which are expressed as follows:

[0053]

[0054] The total DT delay time should not exceed the maximum delay time t kmax [n], which is subject to the following constraints:

[0055] t k [n]≤t kmax [n]

[0056] The production personnel-related information is connected with the manufacturing execution system, enterprise resource planning system and workstation terminal through a standard data interface, and real-time personnel basic archives, scheduling plans, on-duty states, workstation binding and production task execution progress and other structured data are obtained. After extraction, these data are encapsulated as standardized personnel state events with space-time labels in the edge server, and are uploaded to the DT modeling unit through a unified message middleware and sensor data stream synchronization, and then the corresponding state, skill and task attributes in the virtual model are mapped and updated, providing real-time and structured data input for personnel arrangement optimization and production efficiency collaborative prediction.

[0057] Step 3: In the prediction optimization unit, based on the neural network model and machine learning algorithm, the maintenance requirements are predicted according to the equipment operation data, the pre-maintenance plan is made based on the equipment health prediction to optimize the equipment parameters and scheduling sequence; the material remaining situation is predicted according to the in-out warehouse situation, the shortage risk is warned, the material procurement suggestion and distribution strategy are optimized; the influence of environmental parameters on machines and materials is predicted according to real-time temperature and humidity, and a warning is given when the threshold is exceeded; the personnel state is recorded to optimize personnel allocation and scheduling plan; the production scheduling scheme is predicted according to the real-time data of the four monitoring points, and the optimal or near-optimal scheduling is generated under the given constraint conditions by using genetic algorithm for multi-objective simulation and optimization of the production scheduling scheme. The prediction optimization unit sends the prediction results and optimized analysis scheme to the decision-making unit.

[0058] In step 3, in the prediction optimization unit, the workshop environment prediction model, the material inventory prediction model, the machine maintenance prediction and the personnel arrangement model are based on the LSTM neural network model for prediction. The model is trained using the Adam optimization algorithm, which combines momentum gradient descent and adaptive learning rate adjustment, suitable for neural network training of time series data, and adapts to strong time series dependence and trend evolution characteristics. The production scheduling prediction model uses the gradient boosting decision tree model, which is suitable for personnel process matching with relatively clear feature relationships. For the workshop environment prediction model, the input parameters are temperature, humidity, gas concentration and dust concentration, and the output parameters are the environmental parameter changes in the next 15 days; for the material inventory prediction model, the input parameters are historical inventory quantity, daily warehouse-in quantity, daily warehouse-out quantity, production task quantity and environmental temperature and humidity, and the output parameter is the material inventory quantity in the next 15 days; for the machine maintenance prediction model, the input parameters are device vibration frequency, running temperature, running time and historical maintenance records, and the output parameter is the remaining service life; for the personnel arrangement model, the input parameters are personnel attributes, production plan and working time, and the output parameter is personnel process matching; the production scheduling prediction model uses the XGBoost algorithm, which can effectively handle mixed type features and has good interpretability for feature importance. The input data includes historical production task quantity, equipment failure rate, material arrival time, personnel attendance rate, etc., and the output is the production scheduling scheme. During model training, historical and real-time data are uniformly preprocessed, including data integration, data cleaning, feature extraction and sliding window feature construction, to ensure reliable data quality. The initial training of the model mainly uses historical data to establish a preliminary prediction baseline for subsequent combination with real-time data. After the preliminary training is completed, real-time data will be gradually introduced into the training process to make the prediction results consistent with the current actual situation. The rolling training method is adopted, with the latest real-time data in the time window as new samples to balance the training stability and adaptability to new trends.

[0059] Step 4: The decision-making unit generates decisions related to machine equipment, material inventory, environmental temperature and humidity, personnel arrangement and production scheduling scheme. The decisions are generated in combination with the prediction optimization transmitted by the prediction optimization unit, and the decisions are sent to the decision receiving unit of the physical entity layer.

[0060] In step 4, the decision-making unit generates decisions on equipment maintenance, parameter optimization, personnel arrangement and production scheduling based on the processed perception data of the perception device unit and the results of the prediction model for each monitoring point uploaded by the prediction optimization unit. The physical entity layer executes the decisions and feeds back the execution results to the digital twin layer. Through analysis and deduction of the perception data of each monitoring point, resource scheduling instructions, parameter optimization schemes and fault response strategies are generated to guide the production and manufacturing of the furniture factory.

[0061] Step 5: The perception object unit of the physical perception module of the physical entity layer executes the decisions from the decision receiving unit and the optimization suggestions from the prediction optimization unit, and feeds back the execution results to the perception device unit. The perception device unit uploads the feedback data to the digital twin layer for subsequent model updating and optimization.

[0062] The above is only the preferred embodiment of the present application, and the protection scope of the present application is not limited to the above embodiment. Any equivalent modification or change made by a person skilled in the art according to the disclosed content of the present application shall be included in the protection scope recited in the claims.

Claims

1. A digital-twin-based furniture factory production scheduling optimization system, characterized in that: the system comprises a physical entity layer and a digital twin layer; the physical entity layer comprises a physical perception module and a production control module; the physical perception module comprises a perception object unit and a perception device unit; the production control module comprises a decision receiving unit; the digital twin layer comprises a digital twin module; the digital twin module comprises a DT modeling unit, a DT updating unit, a prediction and optimization unit, and a decision making unit; the perception object unit is each monitoring point in the furniture factory environment, the perception device unit is used for real-time state monitoring of each point in the physical scene of the furniture factory, collecting relevant data of the perception object unit, performing multi-source data perception, data preprocessing and standardization, and uploading effective data to the DT modeling unit of the digital twin layer; the DT modeling unit receives data and integrates to obtain multi-dimensional perception data, performs furniture factory modeling and data correlation, establishes a real-time mapping relationship between data and models, i.e. a digital twin model; the DT updating unit updates and corrects the parameters of the digital twin model in a timely manner; the prediction and optimization unit performs production condition prediction and optimization analysis of the furniture factory based on multi-dimensional perception data; the decision making unit formulates production scheduling optimization decisions for the furniture factory according to the prediction and optimization analysis results; the decision receiving unit receives decisions issued by the decision making unit of the digital twin layer and sends them to the perception object unit for execution.

2. The digital-twin-based furniture factory production scheduling optimization system according to claim 1, characterized in that: The perception object unit includes physical scene monitoring points of production machines, production materials, workshop environments, and production personnel, as well as Internet of Things devices and control devices for each monitoring point, which provide monitoring objects for the perception device unit. In addition, the perception object unit is used to execute decisions delivered by the decision receiving unit and feed back the execution results to the perception device unit.

3. The digital-twin-based furniture factory production scheduling optimization system of claim 1, wherein: The perception device unit includes various sensors, identification devices, and state acquisition modules in production equipment, the various sensors include temperature and humidity sensors, dust concentration sensors, gas detectors, pressure sensors, vibration sensors, speed sensors, and current sensors; the identification devices include RFID readers, barcode scanners, face recognition terminals, and cameras; the state acquisition modules in the production equipment include PLC data acquisition modules and frequency converter data acquisition modules.

4. The digital-twin-based furniture factory production scheduling optimization system of claim 1, wherein: The decisions received by the decision receiving unit include production scheduling adjustments, production equipment, production environment, and production personnel parameter adjustments, which are sent to the perception object unit, and the perception object unit executes the above decisions to realize the regulation of the physical production process.

5. The digital-twin-based furniture factory production scheduling optimization system of claim 1, wherein: In the digital twin module, the DT modeling unit is used to receive and store real-time data uploaded from the perception device unit in the physical entity layer after preprocessing and standardization, integrate multi-dimensional data from different perception devices, perform feature extraction and selection on the integrated data, use the processed multi-dimensional perception data as a data source, construct a geometric model of the furniture factory, associate the processed multi-dimensional perception data with the geometric model, establish a real-time mapping relationship between physical entity data and the virtual furniture factory model, and obtain a digital twin model; The DT updating unit is configured to receive the multi-dimensional perception data processed by the DT modeling unit, update and correct the digital twin model parameters in time, maintain the real-time mapping relationship between the physical entity and the virtual model, and keep the digital twin model and the physical entity in real-time synchronization. The prediction and optimization unit is configured to receive the multi-dimensional perception data processed by the DT modeling unit, predict the future states of the equipment failure rate, the material inventory, the environmental temperature and humidity, the personnel arrangement, and the production schedule based on the neural network model and the machine learning algorithm, and optimize the production process based on the actual production target, including the optimization of the machine equipment parameters, the optimization of the workshop environment parameters, the optimization of the production material inventory, the optimization of the personnel arrangement, and the optimization of the production schedule, and send the above prediction and optimization to the decision making unit. The decision making unit is configured to receive the prediction and optimization delivered by the prediction and optimization unit, generate the decision related to the equipment maintenance, the parameter optimization, the personnel arrangement, and the production schedule of the monitoring points, and send the decision to the decision receiving unit of the physical entity layer to dynamically adjust the parameters of the monitoring points and the production schedule based on the decision.

6. A method of using the digital-twin-based furniture factory production scheduling optimization system according to any one of claims 1-5, characterized in that: The method comprises the following steps: Step S1: Based on the perception data of the four monitoring points of the production machine, the production material, the workshop environment, and the production personnel, the running state data of the factory equipment, the material warehouse-out, warehouse-in, and inventory data, the environmental parameters, and the data of the personnel arrangement and work content are collected by the perception device unit of the physical entity layer, the collected raw data is preliminarily processed, including data preprocessing and standardization, and the processed real-time state information is uploaded to the DT modeling unit of the digital twin layer; Step S2: The DT modeling unit of the digital twin layer receives and stores the real-time state information processed by the perception device unit, integrates the multi-dimensional data from different perception devices, extracts and selects the integrated data, constructs the digital twin models of the machine, the material, the environment, the personnel, and the production schedule in the DT modeling unit, realizes the real-time mapping between the physical entity and the virtual model, and updates the digital twin model in the DT updating unit to keep the digital twin model and the physical entity state in real-time synchronization; Step S3: In the prediction and optimization unit, the maintenance demand is predicted based on the equipment running data, the maintenance warning is generated, the material remaining condition is predicted based on the warehouse-in and warehouse-out condition, the shortage risk is warned, the influence of the environmental parameters on the machine and the material is predicted based on the real-time temperature and humidity, and the warning is given when the threshold is exceeded; The personnel state is recorded to provide data support for the production efficiency prediction, the production schedule scheme is predicted based on the real-time data of the four monitoring points, and the generated maintenance warning, material inventory warning, environmental control warning, production efficiency prediction, and production schedule scheme prediction are sent to the decision making unit; Step S4: The decision related to the machine equipment, the material inventory, the environmental temperature and humidity, the personnel arrangement, and the production schedule scheme is generated in the decision making unit, the decision generated by the prediction and optimization of the prediction and optimization unit is combined to dynamically adjust the parameters of the monitoring points and the production schedule scheme, and the decision is sent to the decision receiving unit of the physical entity layer. Step S5: The perception object unit of the physical entity layer physical perception module executes the decision from the decision receiving unit and feeds back the execution result to the perception device unit, which then uploads the feedback data to the digital twin layer for subsequent model updating and optimization, forming a closed-loop process of perception-modeling-prediction-decision-execution-feedback.

7. The method according to claim 6, wherein: In step S1, real-time machine running state data, material inventory data, temperature, humidity, air quality, and personnel scheduling time and work content are obtained; the perception device unit of the physical entity layer is deployed with temperature and humidity sensors, dust detectors, gas detectors, pressure sensors, vibration sensors, speed sensors, current sensors, RFID readers, and camera perception devices to detect and process the perceived data from the physical environment. All data is completed within a fixed frame length to meet real-time requirements.

8. The method of claim 6, wherein: In step S2, the digital twin model is deployed on the edge server, and different communication interfaces are set to realize the bidirectional communication between the virtual model and the entity scene. Based on the perception data of the perception device unit and the perception object unit, the model of the furniture factory is established, and the parameters of the model are configured according to the relevant state information to simulate the real-time running state, establishing the digital twin model of the furniture factory.

9. The method of claim 6, wherein: In step S3, the LSTM neural network model is used in the prediction and optimization unit to predict each monitoring point. The model is trained using the Adam optimization algorithm, which combines momentum gradient descent and adaptive learning rate adjustment, for neural network training of time series data. For the workshop environment prediction model, the input parameters are temperature and humidity, gas concentration, and dust concentration, and the output parameters are the environmental parameter changes in the next 15 days. For the material inventory prediction model, the input parameters are historical inventory quantity, daily inventory quantity, daily delivery quantity, production task quantity, and environmental temperature and humidity, and the output parameters are the material inventory quantity in the next 15 days. For the machine maintenance prediction model, the input parameters are device vibration frequency, operating temperature, operating time, and historical maintenance records, and the output parameter is the remaining service life. For the personnel arrangement model, the input parameters are personnel attributes, production plans, and working hours, and the output parameter is personnel process matching. The production scheduling prediction model uses the XGBoost algorithm, and the input data includes historical production task quantity, device failure rate, material arrival time, and personnel attendance rate, and the output is the production scheduling scheme. During model training, historical and real-time data are uniformly preprocessed, including data integration, data cleaning, feature extraction, and sliding window feature construction, to ensure reliable data quality. The initial training of the model mainly uses historical data to establish a preliminary prediction baseline for subsequent combination with real-time data. After the preliminary training is completed, real-time data will be gradually introduced into the training process to make the prediction results fit the current actual situation. The rolling training method is used, with the real-time data in the latest time window as new samples to balance the training stability and adaptability to new trends.

10. The method of claim 6, wherein: In step S4, the decision generation unit generates decisions related to equipment maintenance, parameter optimization, personnel allocation, and production scheduling of each monitoring point based on the perception data and the prediction optimization. The physical entity layer executes the decisions and feeds back the execution results to the digital twin layer. Through analysis and deduction of the perception data of each monitoring point, resource scheduling instructions, parameter optimization schemes, and fault response strategies are generated to guide the production and manufacturing of the furniture factory.

Images