Intelligent monitoring method and system for heating and ventilation of thousand-level dust-free workshop based on digital twinning

Abstract

The invention discloses an intelligent monitoring method and system for heating and ventilation of a thousand-level dust-free workshop based on digital twinning. The method specifically comprises the steps that SolidWorks software is used for conducting three-dimensional modeling on heating and ventilation equipment and environment of the thousand-level dust-free workshop; a physical device, an IoT sensor network, a PLC controller and a data transmission module are configured to realize real-time acquisition, processing and transmission of a device operation state and environmental parameters; dynamically optimizing the operation state of the equipment and the environment parameters of the workshop; a three-dimensional visual interface is developed based on an OpenUSD platform, and heating and ventilation equipment and environmental parameters in a workshop are visually presented in a dynamic and real-time mode; an equipment health state curve is generated through a time sequence prediction algorithm, and the equipment health score and trend change are visually displayed, so that a user can grasp the equipment operation state in real time. Accurate control over the temperature and humidity, the cleanliness and the pressure difference in the workshop is achieved, the cleanliness and reliability of the production process are improved, and good adaptability and expansibility are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of heating, ventilation and air conditioning, and particularly to an intelligent monitoring system and method for heating, ventilation and air conditioning in a thousand-class dust-free workshop based on digital twin. Background Art

[0002] With the rapid development of high-tech industries, thousand-class dust-free workshops are widely used in fields with extremely high requirements for environmental cleanliness, such as semiconductor manufacturing, biopharmaceuticals, precision instrument production, and food processing. In these fields, the control of temperature, humidity, cleanliness, and differential pressure in the workshop is crucial, which not only affects the stability of the production process but also directly impacts product quality and production efficiency.

[0003] Traditional heating, ventilation and air conditioning systems in dust-free workshops mainly rely on fixed parameter settings, and the central control system is used to uniformly regulate equipment such as fresh air handling units (MAUs), air handling units (AHUs), fan filter units (FFUs), and chiller units.

[0004] However, this method has the following technical bottlenecks:

[0005] (1) Lack of real-time data dynamic regulation ability: The traditional system monitors environmental parameters (such as temperature, humidity, cleanliness, and differential pressure) relying on fixed value settings and cannot dynamically adjust according to the real-time state of the workshop, resulting in high energy consumption and slow response.

[0006] (2) Difficulty in achieving precise control in multiple regions: The environmental requirements in different regions of the workshop vary, and it is often difficult for the traditional system to independently optimize each region while meeting the overall environmental requirements.

[0007] (3) Insufficient equipment operation efficiency and health management: The traditional system has a relatively single monitoring of the operation status of HVAC equipment, making it difficult to predict and prevent potential failures in a timely manner, resulting in low equipment operation efficiency and high maintenance costs.

[0008] As an emerging intelligent tool, digital twin technology realizes real-time bidirectional mapping between physical devices and virtual systems by constructing virtual models of physical entities. In recent years, digital twin technology has been widely used in fields such as industrial manufacturing, building management, and energy optimization, and its dynamic visualization and intelligent decision-making functions provide new ideas for solving the above technical bottlenecks.

[0009] The HVAC intelligent monitoring system based on digital twin technology combines physical devices with digital twin models, collects real-time data in the workshop through the IoT sensor network, optimizes the device operation strategy using artificial intelligence algorithms, and displays the workshop environment status through a three-dimensional digital twin interface. This system can not only dynamically adjust temperature, humidity, cleanliness, and pressure difference, but also predict equipment operation failures, generate optimization plans, and improve the precise control of the workshop environment and energy utilization efficiency. However, there is still room for improvement in the existing digital twin HVAC systems in the following aspects:

[0010] (1) Modeling efficiency and dynamic update: How to efficiently build a three-dimensional model and achieve real-time dynamic update.

[0011] (2) Complexity of optimization algorithms: It is necessary to further develop multi-objective optimization algorithms and fault prediction algorithms suitable for the dust-free workshop environment.

[0012] (3) Human-computer interaction experience: The design of the user interaction interface in the existing system is relatively single, and the virtual reality (VR) technology is not fully utilized to provide an immersive operation experience. Summary of the Invention

[0013] The purpose of the present invention is to provide a design method for an HVAC intelligent monitoring system in a thousand-class dust-free workshop that can meet the environmental requirements of various complex production scenarios, has high environmental control accuracy, high intelligent fault management, energy conservation and consumption reduction, and can improve the operation efficiency and overall production quality of the dust-free workshop.

[0014] The technical solution to achieve the purpose of the present invention is: An HVAC intelligent monitoring method for a thousand-class dust-free workshop based on digital twin, which is provided with a SolidWorks modeling layer, a system configuration and control layer, an AI algorithm optimization layer, a digital twin and VR interaction layer, and a result prediction and feedback layer. The specific steps are as follows:

[0015] Step 1, SolidWorks modeling layer, use SolidWorks software to perform three-dimensional modeling on the HVAC equipment and environment in the thousand-class dust-free workshop;

[0016] Step 2, system configuration and control layer, configure physical devices, IoT sensor network, PLC controller and data transmission module to achieve real-time acquisition, processing and transmission of equipment operation status and environmental parameters;

[0017] Step 3, AI algorithm optimization layer, combine multi-objective optimization algorithm, time series prediction algorithm and deep learning model to achieve dynamic optimization of equipment operation status and workshop environmental parameters;

[0018] Step 4, digital twin and VR interaction layer, develop a three-dimensional visualization interface based on the OpenUSD platform to intuitively present the HVAC equipment and environmental parameters in the workshop in a dynamic and real-time manner;

[0019] Step 5, Result Prediction and Feedback Layer: Generate the device health status curve through the time series prediction algorithm, visually display the device health score and trend changes, so that users can grasp the device operation status in real time.

[0020] A thousand-class dust-free workshop HVAC intelligent monitoring system based on digital twin, which is used to implement the above-mentioned thousand-class dust-free workshop HVAC intelligent monitoring method based on digital twin. The system includes a SolidWorks modeling layer, a system configuration and control layer, an AI algorithm optimization layer, a digital twin and VR interaction layer, and a result prediction and feedback layer, where:

[0021] SolidWorks Modeling Layer: Use SolidWorks software to perform 3D modeling on the HVAC equipment and environment of the thousand-class dust-free workshop;

[0022] System Configuration and Control Layer: Configure physical devices, IoT sensor networks, PLC controllers, and data transmission modules to achieve real-time collection, processing, and transmission of device operation status and environmental parameters;

[0023] AI Algorithm Optimization Layer: Combine multi-objective optimization algorithms, time series prediction algorithms, and deep learning models to achieve dynamic optimization of device operation status and workshop environmental parameters;

[0024] Digital Twin and VR Interaction Layer: Develop a 3D visualization interface based on the OpenUSD platform to visually present the HVAC equipment and environmental parameters in the workshop in a dynamic and real-time manner;

[0025] Result Prediction and Feedback Layer: Generate the device health status curve through the time series prediction algorithm, visually display the device health score and trend changes, so that users can grasp the device operation status in real time.

[0026] Compared with the prior art, the significant advantages of the present invention are as follows: (1) The present invention realizes the precise control of temperature, humidity, cleanliness and differential pressure in the workshop, ensuring that the environmental parameters are stable within the requirements of Class 1000 dust-free environment, effectively improving the cleanliness and reliability of the production process; (2) Through the multi-objective optimization algorithm, while meeting the environmental requirements, the equipment energy consumption is minimized to achieve an energy-saving and efficient operation strategy; (3) Through the AI algorithm and time series prediction model, the system dynamically optimizes the operation status of the equipment, and predicts faults in advance through deep learning technology, provides proactive maintenance suggestions for the equipment, and reduces the failure rate and downtime; (4) Based on the analysis of the health status curve and the fault propagation path, the system can identify potential equipment faults in advance, issue warnings and mark the affected range, providing support for rapid response and handling; (5) By adopting digital twin and VR technologies, users can monitor the operation status of the equipment in real time in an immersive virtual environment and simulate parameter adjustment, greatly improving the management efficiency and interaction experience; (6) Through the IoT sensor network and high-performance data transmission module, the system realizes the real-time collection and upload of environmental parameters and equipment status. Combined with the dynamic visualization interface of the digital twin platform, it provides intuitive and real-time operation feedback for users; (7) The system has a highly flexible modular design, supports workshop expansion and function adjustment, ensures that the system can be dynamically optimized according to real-time needs, and has good adaptability and scalability. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flowchart of the intelligent monitoring method for Class 1000 dust-free workshop HVAC based on digital twin of the present invention.

[0028] Figure 2 It is a network layout diagram of the intelligent monitoring system for Class 1000 dust-free workshop HVAC of the present invention.

[0029] Figure 3a It is the first part flowchart of the program control of the intelligent monitoring system for Class 1000 dust-free workshop HVAC of the present invention.

[0030] Figure 3b It is the second part flowchart of the program control of the intelligent monitoring system for Class 1000 dust-free workshop HVAC of the present invention.

[0031] Figure 3c It is the third part flowchart of the program control of the intelligent monitoring system for Class 1000 dust-free workshop HVAC of the present invention.

[0032] Figure 4 It is a schematic diagram of the electrical box layout of the intelligent monitoring system for Class 1000 dust-free workshop HVAC of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0033] The present invention will be further described in conjunction with specific embodiments. Among them, the drawings are only for illustrative purposes, showing only schematic diagrams, not physical diagrams, and should not be construed as a limitation of this patent.

[0034] The present invention provides a method for intelligent monitoring of HVAC in a thousand-class clean workshop based on digital twin, which is provided with a SolidWorks modeling layer, a system configuration and control layer, an AI algorithm optimization layer, a digital twin and VR interaction layer, and a result prediction and feedback layer. The specific steps are as follows:

[0035] Step 1, SolidWorks modeling layer: Use SolidWorks software to perform 3D modeling on the HVAC equipment and environment in the thousand-class clean workshop.

[0036] Step 2, system configuration and control layer: Configure physical devices, IoT sensor networks, PLC controllers, and data transmission modules to achieve real-time acquisition, processing, and transmission of equipment operating states and environmental parameters.

[0037] Step 3, AI algorithm optimization layer: Combine multi-objective optimization algorithms, time series prediction algorithms, and deep learning models to achieve dynamic optimization of equipment operating states and workshop environmental parameters.

[0038] Step 4, digital twin and VR interaction layer: Develop a 3D visualization interface based on the OpenUSD platform to intuitively present the HVAC equipment and environmental parameters in the workshop in a dynamic and real-time manner.

[0039] Step 5, result prediction and feedback layer: Generate an equipment health status curve through a time series prediction algorithm, intuitively display the equipment health score and trend changes, so that users can grasp the equipment operating state in real time.

[0040] Further, in the SolidWorks modeling layer described in Step 1, use SolidWorks software to perform 3D modeling on the HVAC equipment and environment in the thousand-class clean workshop, specifically as follows:

[0041] Step 1.1: Based on SolidWorks software, perform 3D modeling on the HVAC equipment and environment in the thousand-class clean workshop, including the appearance and dimensions of the fresh air handling unit (MAU), air handling unit (AHU), fan filter unit (FFU), dry coil fan, chiller, and exhaust system, as well as the physical connection relationships between the equipment, including pipeline layouts, valve positions, and interaction nodes between the equipment, to ensure that the equipment structure and functional logic are clearly visible.

[0042] Step 1.2: Embed the operating parameters of the equipment in the 3D model, directly associate the operating parameters of the equipment with the equipment geometric structure through the modeling layer, and form a dynamically adjustable digital twin foundation to ensure that the model can reflect the actual state of the equipment operation in real time.

[0043] Step 1.3: According to the geometric modeling and physical relationship definitions, achieve an integrated presentation of the entire system logic, mark the physical connection relationships between devices in the model, and support subsequent simulation analysis and optimization; design the logical relationships between modules to ensure that the modeling layer can intuitively display device interactions and the overall operation logic of the system.

[0044] Step 1.4: Through the airflow simulation technology built into SolidWorks software, analyze the air circulation path inside the workshop, predict the pressure difference between the clean area and the buffer area through airflow simulation, and optimize the equipment layout through pressure difference calculation and simulation analysis to make the air quality in the clean area meet the standards of a dust-free workshop.

[0045] The pressure difference gradient calculation formula is:

[0046] P d = P clean - P buffer

[0047] where P clean is the pressure in the clean area, and P buffer is the pressure in the buffer area.

[0048] Step 1.5: Based on the modular modeling method, establish independent geometric models for different functional areas of the workshop. The models of each modular area support dynamic update and adjustment, and can be adjusted when the workshop is expanded or the equipment is upgraded.

[0049] Step 1.6: Seamlessly integrate the model with the digital twin system for realizing VR interaction and real-time monitoring visualization.

[0050] Furthermore, for the system configuration and control layer described in Step 2, configure physical devices, IoT sensor networks, PLC controllers, and data transmission modules to achieve real-time acquisition, processing, and transmission of device operation states and environmental parameters, specifically as follows:

[0051] Step 2.1: During layout, according to the functional requirements of the workshop, through digital twin modeling, utilize the collaborative relationship matrix between devices to optimize the equipment layout and operation mode.

[0052] The collaborative relationship matrix M eq (i, j) is:

[0053]

[0054] where c i,j represents the operational dependence between devices i and j, and M eq (i, j) is used to analyze the collaborative working efficiency of devices, thereby dynamically adjusting the equipment layout and operation priorities.

[0055] Step 2.2: Deploy multi-type IoT sensor networks at key positions in the workshop, including temperature and humidity sensors, cleanliness sensors, differential pressure sensors, and energy consumption sensors, for real-time monitoring, data integration, and dynamic grouping, and optimize them using a weighted clustering algorithm;

[0056] Step 2.3: The system processes the collected environmental parameters and equipment operation data in real time through the PLC controller, and dynamically adjusts the fan wind speed, air supply volume, and cooling water flow according to the changes in environmental parameters to ensure a stable workshop environment;

[0057] Step 2.4: The data transmission module uploads the integrated sensor data to the digital twin system and the central platform; through the switch and the hierarchical data transmission strategy, the data transmission module realizes the following features:

[0058] Priority transmission: Prioritize the processing of abnormal data;

[0059] Multi-protocol support: The data transmission module supports multiple communication protocols and adapts to different device and sensor interfaces;

[0060] During the data upload process, the priority queue scheduling algorithm is adopted to sort the data packets according to the importance of the data, reducing the transmission delay of critical data;

[0061] The calculation formula for the queue weight Wdata is as follows:

[0062] Wdata = α·Turgency + β·Ssize + γ·Rreliability

[0063] Where Turgency represents the urgency of the data; Ssize is the data packet size; Rreliability is the reliability requirement of the data packet; α, β, γ are adjustment coefficients;

[0064] Step 2.5: When the system detects an environmental anomaly, it automatically triggers a response mechanism for dynamic wind speed adjustment, multi-device collaborative optimization, and dynamically adjusts the device operation state;

[0065] Dynamic wind speed adjustment: When the differential pressure is lower than the set value, the FFU fan wind speed automatically increases to maintain a positive pressure environment in the clean area;

[0066] Multi-device collaborative optimization: The system adjusts the operation modes of the fresh air unit and the exhaust fan unit according to the real-time states of the clean area and the buffer area, and at the same time restores the target environmental parameters;

[0067] The system uses a distributed control algorithm to dynamically adjust the device operation state. When the differential pressure decreases, the system adjusts the fan wind speed through the following optimization formula, is the fan wind speed before adjustment, The adjusted fan air velocity is:

[0068]

[0069] where ΔP is the target pressure difference P set and the current pressure difference P current The difference between them, that is, ΔP = P set - P current ; k p 、k i 、K d are the PID gain parameters after dynamic optimization by reinforcement learning;

[0070] The system combines predictive control to adjust the operating parameters of other devices in advance to achieve environmental stability;

[0071] Step 2.6: To ensure the stability of data acquisition and equipment operation, configure electrical modules including UPS and power distribution cabinets. On the basis of the UPS power supply system, introduce a fault prediction model to predict possible faults of the power supply module based on the abnormal fluctuation characteristics of operation data;

[0072] The calculation formula for the abnormal feature A(t) is:

[0073]

[0074] where V set 、V current are the set voltage and the actual voltage respectively; I set 、I current are the set current and the actual current respectively;

[0075] When A(t) exceeds the preset threshold, switch to the standby power supply in advance to ensure the continuity of data acquisition and transmission.

[0076] Furthermore, in step 2.2, the real-time monitoring, data integration and dynamic grouping are carried out and optimized by using the weighted clustering algorithm, specifically as follows:

[0077] Real-time monitoring: The temperature, humidity, cleanliness and pressure difference parameters are collected in real time through sensors to ensure that the environment meets the standard of a thousand-class dust-free workshop;

[0078] Data integration: The sensor data is filtered and integrated by the intelligent aggregation method to filter and integrate the redundant data collected at multiple points, improve the data transmission efficiency and reduce the bandwidth occupancy;

[0079] Dynamic grouping: The sensor network is grouped according to the needs of different regions to ensure that the data in important regions is processed and responded to first;

[0080] The formula for the weighted clustering algorithm is:

[0081]

[0082] where w i is the weight of sensor i, and d(p i , p c ) is the distance from sensor i to the cluster center (p i , p c ).

[0083] Furthermore, in step 2.3, the PLC controller integrates an adaptive algorithm to learn and optimize the control strategy, giving priority to meeting the temperature and humidity stability requirements of important areas;

[0084] The reinforcement learning algorithm is used to optimize the control gain parameters and dynamically learn the optimal control strategy under different scenarios. The formula is:

[0085] u(t) = Q(s, a) = r(s, a) + γa′maxQ(s′, a′)

[0086] where u(t) is the control output; Q(s, a) is the value function of state s and action a; r(s, a) is the immediate reward; γa′max is the discount factor, indicating the importance of long-term benefits;

[0087] Through the reinforcement learning algorithm, the PLC controller can adapt to different working conditions and optimize the control of the wind speed, air supply volume, and water flow of HVAC equipment in real time.

[0088] Furthermore, the AI algorithm optimization layer described in step 3 combines the multi-objective optimization algorithm, time series prediction algorithm, and deep learning model to achieve dynamic optimization of the equipment operating state and workshop environment parameters, as follows:

[0089] Step 3.1, multi-objective optimization algorithm: Based on the workshop environment control objectives, a multi-objective optimization model is constructed, including temperature and humidity regulation, cleanliness compliance, and energy consumption minimization optimization;

[0090] During operation, the environmental data and equipment status parameters of each area are analyzed in real time, and the operating status of the fresh air unit, FFU fan, and chiller equipment is dynamically adjusted to achieve a balance between environmental requirements and energy efficiency;

[0091] The multi-objective optimization process realizes the dynamic adjustment of the global optimal solution by constructing an optimization objective function and comprehensively considering the equipment operating efficiency and workshop environmental requirements;

[0092] The objective function is defined as:

[0093] min F(x) = αE(x) + β|T - T set | + γ|H - H set | + δ|C - Cset |

[0094] Among them, E(x) is the total energy consumption of the system; T is the real-time temperature, and T set is the target temperature; H is the real-time humidity, and H set is the target humidity; C is the real-time cleanliness, and C set is the target cleanliness; α, β, and γ are weight coefficients used to balance the priorities among energy consumption, temperature and humidity, and cleanliness;

[0095] The multi-objective optimization model combines real-time environmental parameters and finds the optimal solution through a fast genetic algorithm based on constrained optimization; the fitness function f(x) of the genetic algorithm is:

[0096]

[0097] Solutions with higher fitness are preferentially selected and retained;

[0098] Step 3.2, Time series prediction algorithm: Use the LSTM and Prophet models to model the equipment status and environmental parameters, analyze the historical operation data of the equipment and the real-time collected status data using the time series prediction algorithm, predict the operation trend of the equipment and the future changes of the workshop environmental parameters, and adjust the equipment operation mode and environmental parameters by predicting the trends of cooling water flow, fan wind speed, cleanliness, temperature and humidity;

[0099] The time series prediction formula is:

[0100] y t+1 = f(y t , h t )

[0101] Among them, y t is the environmental or equipment status at time t, and y t+1 is the environmental or equipment status at time t + 1; h t is the hidden state at time t, learned by the LSTM model; f represents the prediction model;

[0102] The Prophet model makes predictions by decomposing the time series into trend, seasonal, and residual parts:

[0103] y(t) = g(t) + s(t) + h(t) + ∈ t

[0104] Among them, g(t) is the long-term trend; s(t) is the periodic change; h(t) is the holiday effect; ∈ t is the random error;

[0105] Combining the long-term prediction ability of Prophet and the short-term dynamic adaptation ability of LSTM, a multi-modal hybrid prediction model is constructed to realize the prediction of equipment operation trends and the prediction of environmental state changes;

[0106] Step 3.3, Deep learning model: Based on the graph neural network GNN, a correlation model between devices is established to analyze the interaction relationships and dependencies among HVAC devices during operation. Through comprehensive analysis of multi-dimensional data, potential anomalies of the devices are identified and possible fault propagation paths are predicted. According to the analysis results, a device health status score is generated for real-time evaluation of the operation reliability of the devices; GNN can capture the interaction relationships between device nodes and update the node states through the adjacency matrix and feature matrix;

[0107] The node state update formula of GNN is:

[0108] H (k+1) =σ(AH (k) W (k) )

[0109] Where, H (k) is the node feature of the k-th layer, and H (k+1) is the node feature of the k+1-th layer; A is the adjacency matrix representing the connection relationship between devices; W (k) is the weight matrix; σ is the activation function;

[0110] By analyzing the multi-dimensional data of the devices, GNN generates a health status score S health :

[0111]

[0112] Where, R i is the fault risk of device i; T i is the total operation time of device i;

[0113] The graph attention network GAT is introduced to perform weighted analysis on key device nodes;

[0114] Step 3.4, Fault prediction and proactive maintenance: Combining the deep learning model to predict and analyze device faults, generating the propagation paths and influence scopes of potential faults. Through the health score and fault risk level generated by the model, the system recommends specific proactive maintenance plans, including device component replacement suggestions, operation parameter adjustment strategies, and maintenance priority rankings;

[0115] Step 3.5, Dynamic Optimization and Real-time Feedback: Based on prediction and optimization, the AI algorithm optimization layer adjusts the operating parameters of the equipment in real time, including the air supply volume of the fresh air unit, the air speed of the FFU fan, and the cooling water temperature of the chiller, to ensure that the equipment operates in an optimal state. At the same time, the system feeds back the optimized operating state and adjustment suggestions to the digital twin interface for users to refer to or manually intervene.

[0116] The energy-saving optimization suggestions V generated according to the environmental requirements and equipment operating status are as follows: f opt For:

[0117] V f opt = arg minF(V f )

[0118] where V f is the multi-objective function of the impact of air speed on the environment.

[0119] Furthermore, the digital twin and VR interaction layer described in Step 4 develops a three-dimensional visualization interface based on the OpenUSD platform, and intuitively presents the HVAC equipment and environmental parameters in the workshop in a dynamic and real-time manner, as follows:

[0120] Step 4.1, The digital twin and VR interaction layer builds a three-dimensional visualization model relying on the OpenUSD platform, covering the panoramic display of HVAC equipment and environmental parameters in the workshop, and performs real-time dynamic updates and multi-dimensional visualization displays;

[0121] Real-time dynamic update: The digital twin model integrates with the IoT sensor network to receive and update the equipment operating status and environmental parameters in real time, ensuring that the model always reflects the actual operating status of the workshop;

[0122] Multi-dimensional visualization display: The digital twin model supports the dynamic display of multi-dimensional parameters, and intuitively presents the air flow distribution, cleanliness change, and temperature and humidity control effect using graphic elements;

[0123] Step 4.2, Synchronize and update the equipment status and environmental data collected by the IoT sensor network with the three-dimensional visualization operation interface in real time, ensuring that users can grasp the environmental changes and equipment operating conditions in the workshop in real time, so as to achieve management and control. By integrating the IoT network and AI optimization algorithm into the digital twin system, real-time monitoring and predictive analysis functions are realized:

[0124] Real-time data stream mapping: Map the temperature, humidity, cleanliness, and differential pressure data collected by IoT sensors to the three-dimensional digital twin model in real time, and display the parameter changes using the method of dynamic numerical superposition;

[0125] AI Model - Driven Optimization: Utilize the prediction results of AI algorithms to dynamically adjust the display of device operating status on the digital twin interface;

[0126] The real - time mapping formula converts the sensor data S i (t) into the dynamic display value M i (t) of the digital twin model:

[0127]

[0128] where S i (t) is the sensor acquisition value at time t; P min and S max are the minimum and maximum values of the sensor data, used for normalized display;

[0129] Step 4.3: Combine with virtual reality (VR) technology, enabling users to achieve real - time monitoring, simulation operation and debugging, and virtual fault drill functions in a virtual environment:

[0130] Real - time monitoring: Users view the operating status and environmental parameters of the device through the VR interface;

[0131] Simulation operation and debugging: Users adjust the device operating parameters and observe the impact of these adjustments on the environment in real - time, providing intuitive feedback;

[0132] Virtual fault drill: Simulate device fault scenarios in the VR environment, analyze the fault propagation path, and generate fault handling suggestions through an interactive visualization interface;

[0133] Adopt motion capture and user behavior analysis functions. By recording the operation behavior of users in the VR environment in real - time, optimize the interface design and interaction efficiency, and evaluate the effectiveness of users' parameter adjustments. The formula is:

[0134]

[0135] where P i adjusted is the parameter adjusted by the user; P i optimal is the optimal parameter recommended by AI; N is the number of adjustments;

[0136] The digital twin and VR interaction layer optimizes the user experience and improves operation efficiency by analyzing the change trend of E eff ;

[0137] Step 4.4: Integrate data from different regions and devices in the digital twin interface to generate a multi - scene heat map of the workshop environment, dynamically displaying the temperature, humidity, cleanliness, and air flow distribution of each region;

[0138] Step 4.5: The digital twin and VR interaction layer generates targeted optimization suggestions through real-time data analysis and user operation feedback. When areas with insufficient cleanliness are found in the VR environment, the system automatically generates the following optimization solutions:

[0139] (1) Adjust the fan speed or air supply volume;

[0140] (2) Improve the operating efficiency of the air conditioning unit;

[0141] (3) Optimize the air flow distribution direction to reduce the risk of pollutant diffusion;

[0142] The feedback control formula is:

[0143] u(t + 1) = u(t) + K·ΔS

[0144] where u(t + 1) is the adjusted equipment parameter; ΔS = S set - S current is the deviation between the set value and the current value; K is the feedback gain coefficient, which adjusts the optimization intensity according to the environmental response.

[0145] Furthermore, in Step 4.4, multi-scenario thermal maps of the workshop environment are generated, specifically including:

[0146] Thermal map generation: The workshop is divided into multiple small units through grid modeling technology, and the parameter values of each unit are calculated in real time through sensor data;

[0147] Abnormal highlighting: When the parameters of a certain area deviate from the set value, the thermal map automatically highlights the abnormal area to assist the user in locating the problem;

[0148] The thermal map generation formula is:

[0149]

[0150] where H(x, y, t) is the environmental parameter value at the position (x, y, t) at time t; S i (x, y, t) is the collected value of sensor i at this position; w i is the weight of sensor i, set based on distance or data credibility.

[0151] Furthermore, in the result prediction and feedback layer described in Step 5, the equipment health status curve is generated through time series prediction algorithms, intuitively showing the equipment health score and trend changes, enabling users to grasp the equipment operation status in real time, specifically as follows:

[0152] Step 5.1, Equipment Health Status Assessment: The result prediction and feedback layer first comprehensively analyzes the historical data and real-time data of equipment operation based on the time series prediction algorithms LSTM and Prophet, generates the equipment health status curve, analyzes and displays the health score and change trend of the equipment. Users can grasp the equipment operation situation in real time through the status curve and identify in advance the equipment with restored operation efficiency or potential faults; finally, generate the equipment health status curve and calculate the health score;

[0153] Health Status Curve Generation: The system uses historical data and real-time monitoring data to predict the future operation status of the equipment, ensuring that users can intuitively grasp the change trend of equipment health;

[0154] Health Score Model: Calculate the health score of the equipment based on the deviation degree between the operation data and the set threshold. The formula is:

[0155]

[0156] where M i is the real-time monitoring value of equipment operation; M optimal is the optimal operation parameter of the equipment; N is the number of monitoring points within the time window;

[0157] When the S health value is lower than the set threshold, the system triggers an alarm and recommends proactive maintenance;

[0158] By combining the short-term dynamic learning ability of LSTM and the long-term trend decomposition model of Prophet, achieve the dynamic balance between short-term early warning and long-term maintenance plan;

[0159] Step 5.2, Fault Propagation Path Prediction: Use the graph neural network GNN to establish a dependency relationship model between equipment, analyze the possible fault propagation paths of equipment, predict the fault impact range, and the system helps users identify the fault propagation paths or environmental areas by marking the potential fault impact areas, providing reference for fault handling;

[0160] Step 5.3, Generation of Operation Optimization Suggestions: Through the multi-objective optimization algorithm, according to the requirements of the workshop environment and the operation status of the equipment, dynamically generate optimization suggestions for equipment operation, adjust the sub-wind speed, cooling water flow or the air supply volume of the new fan group according to real-time data, ensure that the equipment operates in the best state, and at the same time meet the workshop environment control objectives;

[0161] Build a multi-objective optimization model to minimize energy consumption while meeting the environmental requirements of temperature, humidity and cleanliness; The objective function is:

[0162] minE=αE fan +βE cooling +γE ventilation

[0163] Among them, E fan is the energy consumption of the fan, which is related to the wind speed; E cooling is the energy consumption of the cooling system, which is related to the cooling water flow rate; E ventilation is the energy consumption of the fresh air unit, which is related to the air supply volume; α, β, and γ are weight coefficients used to balance the importance of different energy consumption targets;

[0164] The constraint conditions are as follows:

[0165] Temperature constraint: T min ≤T room ≤T max

[0166] Humidity constraint: H min ≤H room ≤H max

[0167] Cleanliness constraint: C room ≥C target

[0168] Differential pressure constraint: ΔP≥ΔP target

[0169] The dynamic feedback control formula optimizes the equipment operation mode according to the real-time feedback;

[0170] The improved dynamic feedback control formula is as follows:

[0171]

[0172] Among them, V f opt is the optimized wind speed of the fan; e(t) is the deviation between the target value and the actual value; K p 、K i 、K d are the gain parameters after dynamic optimization;

[0173] Step 5.4, Operation prediction and energy-saving analysis: Combining the optimization suggestions and the health status assessment, the operation prediction and evaluation module evaluates the best operation state and potential risks of the equipment, and provides an energy-saving plan according to the current status;

[0174] Step 5.5, Active maintenance and feedback plan: According to the health status curve and the fault prediction results, the system generates a maintenance plan, including equipment replacement suggestions and operation parameter adjustment strategies. The maintenance plan is real-time fed back to the user through the digital twin interface, as follows:

[0175] Step 5.5.1, The health status curve is based on time series data, combines the real-time monitoring values and historical operation data, and generates the change trend of the equipment health score over time for identifying potential faults;

[0176] The health score calculation formula is as follows:

[0177]

[0178] Among them, S health (t) is the health score of the device at time t; x current (t) is the current operating parameter of the device; x optimal is the optimal operating parameter of the device;

[0179] When S health (t) is lower than the threshold, the system triggers an alarm and recommends maintenance;

[0180] Step 5.5.2, Time series prediction model:

[0181] Use the LSTM model to predict the future health status curve:

[0182]

[0183] Among them, X t is other relevant environmental parameters, including temperature, humidity and cleanliness;

[0184] Fault prediction is based on the dependency relationship and historical data between devices, generates a fault propagation path and marks the influence range;

[0185] Step 5.5.3, Fault propagation path prediction model:

[0186] Based on the graph neural network GNN, construct a correlation graph between devices:

[0187] H (l+1) =σ(AH (l) W (l) )

[0188] Among them, H (l+1) is the node feature matrix of the l-th layer, that is, the device state; A is the adjacency matrix, indicating the connection relationship between devices; W (l) is the weight matrix; σ is the activation function;

[0189] Through the fault propagation path, analyze the dependency and potential influence area between devices:

[0190]

[0191] Among them, R fault is the propagation risk; d i is the path length from the fault source to device i;

[0192] Generate a maintenance plan According to the health status score and the fault propagation path, generate a specific proactive maintenance plan, including:

[0193] Step 5.5.4, Equipment Replacement Recommendation: Identify the equipment that needs maintenance based on the health status scoring curve and recommend component replacements. The formula for the equipment that needs maintenance is:

[0194] C replace ={i | S health (i) < S threshold}

[0195] where C replace is the set of equipment that needs to be replaced;

[0196] Step 5.5.5, Operating Parameter Adjustment Strategy: Adjust the operating parameters of the equipment. The adjustment formula is:

[0197]

[0198] where α is the adjustment step size; is the gradient of the equipment energy consumption with respect to the operating parameters;

[0199] The maintenance plan is fed back in real time through the digital twin interface. The user can view the equipment status and fault areas in the virtual environment and operate the adjustment of the operating parameters;

[0200] The real-time feedback formula is:

[0201] V feedback = φ(S health , R fault )

[0202] where V feedback is the feedback content; φ is a mapping function used to convert the health score and fault risk into recommendations;

[0203] The feedback includes: highlighting the fault location and the affected area;

[0204] The dynamic display of the maintenance priority is:

[0205] P priority = λ 1 · R fault + λ 2 · (1 - S health )

[0206] Step 5.6, Real-time Data Closed-loop Update: The system continuously updates the health assessment, fault prediction, and operation optimization models through the sensor status data collected in real time, optimizes the later operation data, and synchronously feeds back the adjustment plan to the digital twin interface to form a closed-loop management mechanism;

[0207] Combined with the real-time data, the system forms a closed-loop control mechanism to continuously optimize the operation and maintenance;

[0208] The closed-loop control formula is as follows:

[0209]

[0210] Where Δx is the feedback adjustment amount; β is the feedback gain;

[0211] Finally, the adjustment result is compared with the actual state to generate the next optimization plan.

[0212] The present invention also provides a digital-twin-based intelligent monitoring system for the HVAC of a thousand-class clean workshop, which is used to implement the digital-twin-based intelligent monitoring method for the HVAC of a thousand-class clean workshop. The system includes a SolidWorks modeling layer, a system configuration and control layer, an AI algorithm optimization layer, a digital twin and VR interaction layer, and a result prediction and feedback layer, where:

[0213] SolidWorks modeling layer: Use SolidWorks software to perform 3D modeling on the HVAC equipment and environment of a thousand-class clean workshop;

[0214] System configuration and control layer: Configure physical devices, IoT sensor networks, PLC controllers, and data transmission modules to realize the real-time acquisition, processing, and transmission of equipment operating states and environmental parameters;

[0215] AI algorithm optimization layer: Combine multi-objective optimization algorithms, time series prediction algorithms, and deep learning models to realize the dynamic optimization of equipment operating states and workshop environmental parameters;

[0216] Digital twin and VR interaction layer: Develop a 3D visualization interface based on the OpenUSD platform to intuitively present the HVAC equipment and environmental parameters in the workshop in a dynamic and real-time manner;

[0217] Result prediction and feedback layer: Generate a curve of the equipment health state through a time series prediction algorithm, intuitively display the equipment health score and trend changes, so that users can grasp the equipment operating state in real time.

[0218] In all the examples shown and discussed here, any specific value should be construed as merely exemplary and not as a limitation.

[0219] Embodiment

[0220] This embodiment provides a digital-twin-based intelligent monitoring system for the HVAC of a thousand-class clean workshop. By combining 3D modeling, Internet of Things (IoT), artificial intelligence (AI) optimization algorithms, and virtual reality (VR) technology, it realizes the real-time monitoring and precise adjustment control of the temperature, humidity, cleanliness, and pressure difference in the clean workshop, and solves the problems of slow response, high power consumption, and difficulty in predicting equipment failures in the existing system.

[0221] CombinationFigure 1 , Figure 2 , Figures 3a to 3c and Figure 4 , a design method of a thousand - level dust - free workshop HVAC intelligent monitoring system based on digital twin includes the following steps:

[0222] Step 1: The SolidWorks modeling layer is the core foundation of this system. Precise 3D modeling of the HVAC equipment and environment in the thousand - level dust - free workshop is carried out through SolidWorks software. This modeling layer comprehensively covers the key components of the HVAC system, including equipment such as the makeup air unit (MAU), air handling unit (AHU), fan filter unit (FFU), dry coil fan, chiller, and exhaust system. The model not only presents the geometric structure of the equipment in detail but also integrates the equipment operation parameters, installation locations, and their physical connection relationships, ensuring the structural visualization and logical integrity of the entire system. In addition, based on airflow simulation technology, this layer can accurately analyze the airflow organization inside the workshop, optimize the equipment layout, reduce the risk of pollutant diffusion between regions, and ensure reasonable pressure difference control between the clean area and the buffer area. Specifically as follows:

[0223] Step 1.1: Based on SolidWorks software, carry out precise 3D modeling of the HVAC equipment and environment in the thousand - level dust - free workshop, including the appearance and dimensions of the makeup air unit (MAU), air handling unit (AHU), fan filter unit (FFU), dry coil fan, chiller, exhaust system, and the physical connection relationships between the equipment, including pipeline layout, valve positions, and interaction nodes between the equipment, ensuring that the equipment structure and functional logic are clearly visible;

[0224] Step 1.2: Embed the operation parameters of the equipment in the 3D model, such as the wind speed and air supply volume of the fan, the cooling capacity of the chiller, and the temperature and humidity range of air circulation. Directly associate these parameters with the equipment geometric structure through the modeling layer to form a dynamically adjustable digital twin foundation, ensuring that the model can reflect the actual state of equipment operation in real - time;

[0225] Step 1.3: According to the geometric modeling and physical relationship definitions, achieve an integrated presentation of the entire system logic. Mark the physical connection relationships between the equipment in the model to support subsequent simulation analysis and optimization; reasonably design the logical relationships between each module to ensure that the modeling layer can intuitively display equipment interaction and the overall operation logic of the system;

[0226] Step 1.4: Through the airflow simulation technology built into SolidWorks software, analyze the air circulation path inside the workshop. Predict the pressure difference between the clean area and the buffer area through airflow simulation. Through pressure difference calculation and simulation analysis, optimize the equipment layout to reduce the risk of pollutant diffusion between regions and ensure that the air quality in the clean area meets the dust - free workshop standards;

[0227] The calculation formula for the pressure difference gradient is as follows:

[0228] P d = P clean - P buffer

[0229] Where P clean is the pressure in the clean area, and P buffer is the pressure in the buffer area;

[0230] Step 1.5: Based on the modular modeling method, establish independent geometric models for different functional areas of the workshop, such as the cutting workshop, assembly workshop, aging workshop, etc.; the models of each modular area support dynamic update and adjustment, facilitating quick adjustment during workshop expansion or equipment upgrade, and enhancing the flexibility and scalability of the system;

[0231] Step 1.6: Seamlessly integrate the model with the digital twin system, providing a visualization foundation for subsequent VR interaction and real-time monitoring.

[0232] The SolidWorks modeling layer enables users to intuitively understand the structure and operating status of the workshop and equipment through detailed three-dimensional display, providing strong support for the optimization of the HVAC system.

[0233] Step 2: The system configuration and control layer configures physical devices, IoT sensor networks, PLC controllers, and data transmission modules to achieve real-time acquisition, processing, and transmission of equipment operating status and environmental parameters. The physical devices cover core devices such as the makeup air unit (MAU), air handling unit (AHU), fan filter unit (FFU), dry coil fan, and chiller. The IoT sensor network uses temperature and humidity sensors, cleanliness sensors, differential pressure sensors, and energy consumption sensors arranged at key positions to monitor multi-dimensional environmental parameters and equipment status in the workshop in real time. Through the PLC controller and data transmission module, the collected sensor data is efficiently integrated and uploaded to the digital twin system. The data transmission module transmits the operation data to the central platform through a high-performance switch to ensure that the system can respond to environmental changes and equipment anomalies in real time, as follows:

[0234] Step 2.1: For core physical devices such as the makeup air unit, air handling unit, fan filter unit, dry coil fan, and chiller, consider the dynamic optimization of workshop function requirements during layout. According to the workshop function requirements, through digital twin modeling, utilize the collaborative relationship matrix M eq of the equipment to optimize the equipment layout and operation mode;

[0235] The collaborative relationship matrix between the equipment is as follows:

[0236]

[0237] Where Ci,j Represents the operational dependency between devices i and j. The matrix is used to analyze the collaborative working efficiency of devices, thereby dynamically adjusting the device layout and operation priorities;

[0238] Step 2.2: Arrange multi-type IoT sensor networks at key positions in the workshop, including temperature and humidity sensors, cleanliness sensors, differential pressure sensors, and energy consumption sensors, for real-time monitoring, data integration, and dynamic grouping, and optimize using the weighted clustering algorithm;

[0239] Real-time monitoring: Real-time collect temperature and humidity, cleanliness, and differential pressure parameters through sensors to ensure that the environment meets the standards of a thousand-class dust-free workshop;

[0240] Data integration: Sensor data is filtered and integrated for redundant data collected from multiple points through intelligent aggregation, improving data transmission efficiency and reducing bandwidth occupancy;

[0241] Dynamic grouping: The sensor network is grouped according to the requirements of different regions to ensure that data in important regions is processed and responded to first;

[0242] The formula for the weighted clustering algorithm is:

[0243]

[0244] where w i is the weight of sensor i, and d(p i , p c ) is the distance from sensor i to the clustering center (p i , p c );

[0245] Step 2.3: The system processes the collected environmental parameters and device operation data in real time through the PLC controller, and dynamically adjusts the fan wind speed, air supply volume, and cooling water flow according to the changes in environmental parameters to ensure the stability of the workshop environment;

[0246] The PLC controller integrates an adaptive algorithm, which can learn and optimize control strategies, prioritize meeting the temperature and humidity stability requirements of important regions, and reduce energy consumption at the same time;

[0247] Use the reinforcement learning algorithm to optimize the control gain parameters and dynamically learn the best control strategies in different scenarios. The formula is:

[0248] u(t) = Q(s,a) = r(s,a) + γa′maxQ(s′,a′)

[0249] where u(t) is the control output; Q(s,a) is the value function of state s and action a; r(s,a) is the immediate reward; γa′max is the discount factor, indicating the importance of long-term benefits;

[0250] Through this algorithm, the PLC controller can adapt to different working conditions and optimize the control of the air speed, air supply volume, and water flow of HVAC equipment in real time;

[0251] Step 2.4: The data transmission module uploads the integrated sensor data to the digital twin system and the central platform; through a high-performance switch and a hierarchical data transmission strategy, the data transmission module achieves the following features:

[0252] Priority transmission: Prioritize the processing of abnormal data to ensure the system's quick response;

[0253] Multi-protocol support: The data transmission module supports multiple communication protocols, adapts to different device and sensor interfaces, and enhances compatibility;

[0254] During the data upload process, the priority queue scheduling algorithm is adopted to sort the data packets according to their importance, reducing the transmission delay of critical data;

[0255] Queue weight calculation formula:

[0256] Wdata = α·Turgency + β·Ssize + γ·Rreliability

[0257] Where Turgency represents the urgency of the data; Ssize is the data packet size; Rreliability is the reliability requirement of the data packet; α, β, γ are adjustment coefficients;

[0258] Step 2.5: When the system detects an environmental anomaly, such as the differential pressure being lower than the set value or the cleanliness decreasing, it automatically triggers a response mechanism to perform dynamic air speed adjustment, multi-device collaborative optimization, and dynamically adjust the operating state of the equipment;

[0259] Dynamic air speed adjustment: When the differential pressure is lower than the set value, the air speed of the FFU fan automatically increases to ensure a positive pressure environment in the clean area;

[0260] Multi-device collaborative optimization: The system adjusts the operating modes of the fresh air unit and the exhaust fan unit according to the real-time states of the clean area and the buffer area, avoiding energy waste and quickly restoring the target environmental parameters;

[0261] The system adopts an improved distributed control algorithm to dynamically adjust the operating state of the equipment. When the differential pressure decreases, the system adjusts the air speed Vf of the fan through the following optimization formula:

[0262]

[0263] Where ΔP = P set -P currentis the difference between the target pressure difference and the current pressure difference; Kp, Ki, and Kd are the PID gain parameters optimized dynamically by reinforcement learning;

[0264] The system combines predictive control to adjust in advance the operating parameters such as the air supply volume and the cooling water flow of other devices to achieve the rapid stability of the environment;

[0265] Step 2.6: To ensure the stability of data acquisition and equipment operation, configure the UPS, power distribution cabinet, and other electrical modules. The UPS system can provide continuous power support during power failures to ensure the uninterrupted operation of the system; on the basis of the traditional UPS power supply system, introduce a fault prediction model to predict the possible faults of the power supply module based on the abnormal fluctuation characteristics of the operation data; the abnormal feature calculation formula is:

[0266]

[0267] where, V set and V current are the set voltage and the actual voltage respectively; I set and I current are the set current and the actual current respectively;

[0268] When A(t) exceeds the preset threshold, the system switches to the standby power supply in advance to ensure the continuity of data acquisition and transmission.

[0269] Step 3: The AI algorithm optimization layer combines multi-objective optimization algorithms, time series prediction algorithms, and deep learning models to achieve dynamic optimization of the equipment operating status and workshop environment parameters. The multi-objective optimization algorithm is responsible for finding the best balance among temperature and humidity control, cleanliness compliance, and energy consumption minimization to ensure that the workshop environment meets the requirements of a thousand-level dust-free environment while reducing energy consumption. Time series prediction algorithms such as LSTM and Prophet models are based on historical data and real-time collected data to predict the future changes in equipment operating trends and environmental parameters, providing a reliable basis for fault warning and environmental control strategies. At the same time, deep learning models such as the graph neural network GNN can deeply analyze the correlation between devices and the fault propagation path. Through the comprehensive analysis of multi-dimensional data such as vibration and energy consumption, potential faults of the equipment can be discovered in advance, a health status score can be generated, and an active maintenance plan can be recommended, specifically as follows:

[0270] Step 3.1: Multi-objective optimization algorithm: Based on the workshop environment control objectives, construct a multi-objective optimization model, including temperature and humidity control, cleanliness compliance, and energy consumption minimization optimization; during operation, the system dynamically adjusts the operating status of the fresh air unit, FFU fan, and chiller equipment by real-time analyzing the environmental data such as temperature, humidity, cleanliness, and pressure difference in each area and the equipment status parameters to achieve the best balance between environmental requirements and energy efficiency;

[0271] The multi-objective optimization process realizes the dynamic adjustment of the global optimal solution by constructing an optimization objective function and comprehensively considering the equipment operation efficiency and the workshop environment requirements; the objective function is defined as:

[0272] minF(x)=αE(x)+β|T - T set |+γ|H - H set |+δ|C - C set |

[0273] where E(x) is the total energy consumption of the system; T is the real-time temperature, and T set is the target temperature; H is the real-time humidity, and H set is the target humidity; C is the real-time cleanliness, and C set is the target cleanliness; α, β, and γ are weight coefficients used to balance the priorities among energy consumption, temperature and humidity, and cleanliness;

[0274] The multi-objective optimization model combines the real-time environmental parameters and finds the optimal solution through a fast genetic algorithm based on constrained optimization; the fitness function of the genetic algorithm is:

[0275]

[0276] Solutions with higher fitness are preferentially selected and retained to improve the efficiency of the optimization process;

[0277] Step 3.2, Time series prediction algorithm: The LSTM and Prophet models are used to accurately model the equipment status and environmental parameters. The time series prediction algorithm is used to analyze the historical operation data of the equipment and the real-time collected status data, predict the operation trend of the equipment and the future changes of the workshop environmental parameters. By predicting the trends of cooling water flow, fan wind speed, cleanliness, temperature and humidity, it provides a scientific basis for adjusting the equipment operation mode and environmental parameters in advance to ensure the stability of the system operation;

[0278] The time series prediction formula is:

[0279] y t+1 =f(y t ,h t )

[0280] where y t is the environmental or equipment status at time t; h t is the hidden state at time t, learned by the LSTM model; f represents the prediction model;

[0281] The Prophet model predicts by decomposing the time series into trend, seasonal, and residual parts:

[0282] y(t)=g(t)+s(t)+h(t)+∈ t

[0283] Among them, g(t) is the long-term trend; s(t) is the periodic change; h(t) is the holiday effect; ∈ t is the random error;

[0284] Combining the long-term prediction ability of Prophet and the short-term dynamic adaptation ability of LSTM, a multi-modal hybrid prediction model is constructed to achieve high-precision prediction of future equipment operation trends and environmental state changes;

[0285] Step 3.3, Deep learning model: Based on the graph neural network GNN, a correlation model between devices is established to deeply analyze the interaction relationship and dependence of HVAC equipment during operation. Through comprehensive analysis of multi-dimensional data such as vibration, energy consumption, and operating temperature, potential anomalies of the equipment are identified and possible fault propagation paths are predicted. The system generates a device health status score based on the analysis results for real-time evaluation of the operation reliability of the equipment; GNN can capture the complex interaction relationships between device nodes and update the node status through the adjacency matrix and feature matrix;

[0286] The node status update formula of GNN is:

[0287] H (k+1) = σ(AH (k) W (k) )

[0288] Among them, H (k) is the node feature of the k-th layer; A is the adjacency matrix representing the connection relationship between devices; W (k) is the weight matrix; σ is the activation function;

[0289] By analyzing the multi-dimensional data of the equipment, GNN generates a health status score S health :

[0290]

[0291] Among them, R i is the failure risk of device i; T i is the total operating time of device i;

[0292] The graph attention network GAT is introduced to perform weighted analysis on key device nodes, improving the accuracy of fault prediction and the interpretability of fault propagation paths;

[0293] Step 3.4, Fault prediction and proactive maintenance: Combining the deep learning model to predict and analyze equipment faults, generating the propagation path and influence range of potential faults. Through the health score and fault risk level generated by the model, the system recommends specific proactive maintenance plans, including equipment component replacement suggestions, operating parameter adjustment strategies, and maintenance priority rankings, thereby reducing the risk of fault downtime;

[0294] Step 3.5, Dynamic Optimization and Real-time Feedback: Based on prediction and optimization, the AI algorithm optimization layer adjusts the operating parameters of the equipment in real time, including the air supply volume of the fresh air unit, the wind speed of the FFU fan, and the cooling water temperature of the chiller, to ensure that the equipment operates in an optimal state. At the same time, the system feeds back the optimized operating state and adjustment suggestions to the digital twin interface for users to refer to or manually intervene;

[0295] Regarding the environmental requirements and the operating state of the equipment, the energy-saving optimization suggestions generated by the system are:

[0296]

[0297] where V f is the multi-objective function of the wind speed's impact on the environment;

[0298] By adjusting the wind speed of the fan and the cooling water flow in real time, intelligent management of equipment energy consumption is achieved.

[0299] Step 4, The digital twin and VR interaction layer develops a three-dimensional visualization interface based on the OpenUSD platform, presenting the HVAC equipment and environmental parameters in the workshop in a dynamic and real-time manner. Through deep integration with the IoT sensor network and AI algorithms, the interface can dynamically display key parameters such as the operating state of the equipment, temperature and humidity, cleanliness, pressure difference, and air flow distribution, helping users to understand the workshop environment and the operating conditions of the equipment in real time. Combining virtual reality (VR) technology, this interaction layer further enhances the user experience, enabling users to monitor, operate, and debug the equipment in an immersive virtual environment. Users can simulate adjustments to the equipment operating parameters through VR, such as modifying the wind speed of the fan or the supply air temperature, and observe in real time the impact of these adjustments on the workshop environment, as follows:

[0300] Step 4.1, The digital twin and VR interaction layer builds a high-precision three-dimensional visualization model based on the OpenUSD platform, covering the panoramic display of HVAC equipment such as MAU, AHU, FFU, etc. and environmental parameters such as temperature and humidity, cleanliness, pressure difference, and air flow distribution in the workshop, and performs real-time dynamic updates and multi-dimensional visualization displays;

[0301] Real-time dynamic update: The digital twin model integrates with the IoT sensor network to receive and update the equipment operating state and environmental parameters in real time, ensuring that the model always reflects the actual operating state of the workshop;

[0302] Multi-dimensional visualization display: The digital twin model supports the dynamic display of multi-dimensional parameters, using graphical elements to visually present the air flow distribution, cleanliness changes, and temperature and humidity control effects;

[0303] Step 4.2: Synchronize and update the device status and environmental data collected by the Internet of Things sensor network with the three-dimensional visualization operation interface in real time to ensure that users can quickly grasp the environmental changes and efficient device operation status in the workshop, so as to achieve management and control. By integrating the IoT network and AI optimization algorithms into the digital twin system, real-time monitoring and predictive analysis functions are realized:

[0304] Real-time mapping of data flow: Real-time map the temperature, humidity, cleanliness, and differential pressure data collected by IoT sensors into the three-dimensional digital twin model, and display the parameter changes in the way of dynamic numerical superposition;

[0305] AI model-driven optimization: Utilize the prediction results of AI algorithms to dynamically adjust the display of the device operation status on the digital twin interface;

[0306] The real-time mapping formula converts the sensor data S i (t) into the dynamic display value M i (t) of the digital twin model:

[0307]

[0308] where S i (t) is the sensor acquisition value at time t; S min and S max are the minimum and maximum values of the sensor data, which are used for normalized display;

[0309] Step 4.3: Combine virtual reality (VR) technology to provide users with an immersive operation and monitoring experience. Users can achieve real-time monitoring, simulation operation and debugging, and virtual fault drill functions in the virtual environment:

[0310] Real-time monitoring: Users can view the operation status and environmental parameters of the device through the VR interface;

[0311] Simulation operation and debugging: Users can adjust the device operation parameters and observe the impact of these adjustments on the environment in real time, providing intuitive feedback;

[0312] Virtual fault drill: Simulate device fault scenarios in the VR environment, analyze the fault propagation path, and generate fault handling suggestions through an interactive visualization interface;

[0313] Adopt motion capture and user behavior analysis functions. By recording the operation behavior of users in the VR environment in real time, optimize the interface design and interaction efficiency, and evaluate the effectiveness of users' parameter adjustments. The formula is:

[0314]

[0315] where are the parameters adjusted by the user; Optimal parameters recommended for AI; N is the number of adjustments;

[0316] The digital twin and VR interaction layer optimizes the user experience and improves operation efficiency by analyzing the eff changing trend of E;

[0317] Step 4.4: Integrate data from different regions and devices in the digital twin interface to generate a multi-scenario heat map of the workshop environment, dynamically displaying the temperature, humidity, cleanliness, and air flow distribution in each region;

[0318] Heat map generation: The workshop is divided into multiple small units through grid modeling technology, and the parameter values of each unit are calculated in real time through sensor data;

[0319] Abnormal highlighting: When the parameters of a certain region deviate from the set value, the heat map automatically highlights and marks the abnormal region to assist the user in quickly locating the problem;

[0320] The heat map generation formula is:

[0321]

[0322] Among them, H(x, y, t) is the environmental parameter value at position (x, y, t) at time t; S i (x, y, t) is the acquisition value of sensor i at this position; w i is the weight of sensor i, set based on distance or data credibility;

[0323] Step 4.5: The digital twin and VR interaction layer generate targeted optimization suggestions through real-time data analysis and user operation feedback. When a region with insufficient cleanliness is found in the VR environment, the system automatically generates the following optimization plan:

[0324] (1) Adjust the fan speed or air supply volume;

[0325] (2) Improve the operating efficiency of the air conditioning unit;

[0326] (3) Optimize the air flow distribution direction to reduce the risk of pollutant diffusion;

[0327] The feedback control formula is:

[0328] u(t + 1) = u(t) + K·ΔS

[0329] Among them, u(t + 1) is the adjusted equipment parameter; ΔS = S set -S current is the deviation between the set value and the current value; K is the feedback gain coefficient, which adjusts the optimization intensity according to the environmental response.

[0330] Step 5. The result prediction and feedback layer generates the device health status curve through a time series prediction algorithm, intuitively displays the device health score and trend changes, enabling users to grasp the device operation status in real time. The graph neural network GNN is used to predict the fault propagation path, analyze the dependency relationship between devices, and mark the potential fault impact area, providing a reference for rapid handling. The system generates operation optimization suggestions according to environmental requirements and device status through a multi-objective optimization algorithm, such as adjusting the fan wind speed or optimizing the cooling water flow. Combining these data, the operation prediction module evaluates the optimal operation status and potential risks of the device, and provides energy-saving solutions and maintenance plans for users in real time to ensure the stability of the workshop environment and the efficient operation of the device, as follows:

[0331] Step 5.1. Device health status assessment: The result prediction and feedback layer first comprehensively analyzes the historical data and real-time data of device operation based on the time series prediction algorithms LSTM and Prophet, generates the device health status curve, and analyzes and displays the device health score and its change trend. Users can grasp the device operation situation in real time through the status curve and identify in advance the devices with restored operation efficiency or potential faults; finally, generate the device health status curve and calculate the health score;

[0332] Generation of the health status curve: The system uses historical data and real-time monitoring data to predict the future operation status of the device, ensuring that users can intuitively grasp the trend of device health changes;

[0333] Health score model: Calculate the health score of the device based on the deviation degree between the operation data and the set threshold. The formula is:

[0334]

[0335] where M i is the real-time monitoring value of device operation; M optimal is the optimal operation parameter of the device; N is the number of monitoring points within the time window;

[0336] When the S health value is lower than the set threshold, the system triggers an alarm and recommends proactive maintenance;

[0337] By combining the short-term dynamic learning ability of LSTM and the long-term trend decomposition model of Prophet, the dynamic balance between short-term early warning and long-term maintenance plans is achieved;

[0338] Step 5.2. Prediction of the fault propagation path: Use the graph neural network GNN to establish a dependency relationship model between devices, analyze the possible fault propagation paths of devices, and predict the fault impact range. The system helps users quickly identify the fault propagation path or environmental area by marking the potential fault impact area, providing a reliable reference for fault handling;

[0339] Step 5.3, Run Optimization Suggestion Generation: Through a multi-objective optimization algorithm, according to the environmental requirements such as the temperature, humidity, cleanliness, and differential pressure in the workshop and the operating status of the equipment, dynamically generate optimization suggestions for the equipment operation, and adjust the sub-wind speed, cooling water flow rate, or the supply air volume of the new air handling unit according to the real-time data to ensure that the equipment operates in the best state while meeting the workshop environmental control objectives;

[0340] Construct a multi-objective optimization model to minimize energy consumption while meeting the environmental requirements of temperature, humidity, and cleanliness; The objective function is:

[0341] minE = αE fan +βE cooling +γE ventilation

[0342] where E fan is the energy consumption of the fan, which is related to the wind speed; E cooling is the energy consumption of the cooling system, which is related to the cooling water flow rate; E ventilation is the energy consumption of the new air handling unit, which is related to the supply air volume; α, β, and γ are weight coefficients used to balance the importance of different energy consumption objectives;

[0343] The constraint conditions are:

[0344] Temperature constraint: T min ≤T room ≤T max

[0345] Humidity constraint: H min ≤H room ≤H max

[0346] Cleanliness constraint: C room ≥C target

[0347] Differential pressure constraint: ΔP≥ΔP target

[0348] The dynamic feedback control formula optimizes the equipment operation mode according to the real-time feedback; The improved dynamic feedback control formula is:

[0349]

[0350] where, is the optimized wind speed of the fan; e(t) is the deviation between the target value and the actual value; Kp, Ki, and Kd are the gain parameters after dynamic optimization;

[0351] Step 5.4, Run Prediction and Energy Saving Analysis: Combine the optimization suggestions and health status assessment, run the prediction and evaluation module to determine the optimal operating state and potential risks of the equipment, and provide an energy-saving plan based on the current status; the system can recommend switching the operating mode of the refrigeration controller or optimizing the area scheduling to further improve the operating efficiency;

[0352] Step 5.5, Active Maintenance and Feedback Plan: Based on the health status curve and fault prediction results, the system generates a maintenance plan, including equipment replacement suggestions and operating parameter adjustment strategies. The maintenance plan is real-time fed back to the user through the digital twin interface to ensure that the user can take quick actions to avoid equipment failures or environmental out-of-control situations. Specifically as follows:

[0353] Step 5.5.1, Health Status Curve: Based on time series data, combine the real-time monitoring values with historical operating data to generate the changing trend of the equipment's health score over time, which is used to identify potential faults;

[0354] The formula for calculating the health score is:

[0355]

[0356] where, S health (t) is the health score of the equipment at time t; x current (t) is the current operating parameter of the equipment; x optimal is the optimal operating parameter of the equipment;

[0357] When S health (t) is lower than a certain threshold, such as Sthreshold = 0.7, the system triggers an early warning and recommends maintenance;

[0358] Step 5.5.2, Time Series Prediction Model:

[0359] Use the LSTM model to predict the future health status curve:

[0360]

[0361] where, X t is other relevant environmental parameters, including temperature, humidity, and cleanliness;

[0362] Fault prediction is based on the dependency relationships and historical data among equipment to generate a fault propagation path and mark the impact scope;

[0363] Step 5.5.3, Fault Propagation Path Prediction Model:

[0364] Based on the graph neural network GNN, construct a relevance graph among equipment:

[0365] H (l+1) = σ(AH(l) W (l) )

[0366] Among them, H (l+1) is the node feature matrix of the l-th layer, that is, the device state; A is the adjacency matrix, representing the connection relationship between devices; W (l) is the weight matrix; σ is the activation function;

[0367] Analyze the dependencies and potential impact areas between devices through the fault propagation path:

[0368]

[0369] Among them, R fault is the propagation risk; d i is the path length from the fault source to device i;

[0370] Generate a maintenance plan Based on the health status score and the fault propagation path, generate a specific proactive maintenance plan, including:

[0371] Step 5.5.4, Device replacement recommendation: Identify the devices that need maintenance according to the health status score curve and recommend component replacement. The calculation formula for the devices that need maintenance is:

[0372] C replace ={i|S health (i)<S threshold}

[0373] Among them, C replace is the set of devices that need to be replaced;

[0374] Step 5.5.5, Operating parameter adjustment strategy: Adjust the operating parameters of the device to reduce the operating pressure and avoid further damage. The adjustment formula is:

[0375]

[0376] Among them, α is the adjustment step size; is the gradient of the device energy consumption with respect to the operating parameters;

[0377] Real-time feedback of the maintenance plan through the digital twin interface. Users can view the device status and the fault area in the virtual environment and operate the adjustment of the operating parameters. The real-time feedback formula is:

[0378] V feedback =φ(S health ,R fault )

[0379] Among them, V feedback is the feedback content; φ is the mapping function used to convert the health score and the fault risk into suggestions;

[0380] The feedback includes:

[0381] Highlight the fault location and the affected area;

[0382] Dynamically display the maintenance priority as:

[0383] P priority = λ 1 ·R fault + λ 2 ·(1 - S health )

[0384] Step 5.6, Real-time data closed-loop update: The system continuously updates the health assessment, fault prediction, and operation optimization models through the sensor status data collected in real time to ensure the accuracy and reliability of the prediction results and optimization suggestions. The optimized later operation data and adjustment plan are synchronously fed back to the digital twin interface to form a closed-loop management mechanism;

[0385] Combined with the real-time data, the system forms a closed-loop control mechanism to continuously optimize operation and maintenance;

[0386] The closed-loop control formula is:

[0387]

[0388] where Δx is the feedback adjustment amount; β is the feedback gain;

[0389] The system compares the adjustment result with the actual state to generate the next optimization plan.

[0390] The above is only the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for intelligent HVAC monitoring of Class 1000 cleanrooms based on digital twins, characterized in that: There are SolidWorks modeling layer, system configuration and control layer, AI algorithm optimization layer, digital twin and VR interaction layer, and result prediction and feedback layer. The specific steps are as follows: Step 1: SolidWorks modeling layer: Use SolidWorks software to perform 3D modeling of the HVAC equipment and environment of the Class 1000 clean room; Step 2: System configuration and control layer: configure physical devices, IoT sensor networks, PLC controllers, and data transmission modules to achieve real-time collection, processing, and transmission of equipment operating status and environmental parameters; Step 3: AI algorithm optimization layer combines multi-objective optimization algorithm, time series prediction algorithm and deep learning model to achieve dynamic optimization of equipment operation status and workshop environmental parameters; Step 4: Digital twin and VR interaction layer, develop a three-dimensional visualization interface based on the OpenUSD platform to intuitively present the HVAC equipment and environmental parameters in the workshop in a dynamic and real-time manner; Step 5: Result prediction and feedback layer generates the equipment health status curve through the time series prediction algorithm, intuitively displays the equipment health score and trend changes, and enables users to understand the equipment operation status in real time.

2. The method for intelligent monitoring of HVAC in a Class 1000 clean room based on digital twin according to claim 1 is characterized in that: In the SolidWorks modeling layer described in step 1, the HVAC equipment and environment of the Class 1000 clean room are 3D modeled using SolidWorks software, as follows: Step 1.1: Based on SolidWorks software, 3D modeling of the HVAC equipment and environment in the Class 1000 clean room is performed, including the appearance and size of the fresh air unit MAU, air conditioning unit AHU, fan filter unit FFU, dry coil fan, refrigeration unit, exhaust system, and the physical connection relationship between the equipment, including pipeline layout, valve position and interaction nodes between equipment, to ensure that the equipment structure and functional logic are clearly visible; Step 1.2: Embed the operating parameters of the equipment in the 3D model, and directly associate the operating parameters of the equipment with the equipment geometry through the modeling layer to form a dynamically adjustable digital twin foundation to ensure that the model can reflect the actual status of the equipment operation in real time; Step 1.3: Based on geometric modeling and physical relationship definition, realize the integrated presentation of the entire system logic, mark the physical connection relationship between devices in the model, and support subsequent simulation analysis and optimization; design the logical relationship between modules to ensure that the modeling layer can intuitively display the device interaction and the overall operation logic of the system; Step 1.4: Use the built-in airflow simulation technology of SolidWorks software to analyze the air circulation path inside the workshop, predict the pressure difference between the clean area and the buffer zone through airflow simulation, and optimize the equipment layout through pressure difference calculation and simulation analysis to make the air quality in the clean area meet the dust-free workshop standards; The pressure gradient calculation formula is: P d =P clean -P buffer Where P clean is the pressure of the clean area, P buffer is the pressure of the buffer zone; Step 1.5: Based on the modular modeling method, establish independent geometric models for different functional areas of the workshop. The models of each modular area support dynamic update and adjustment, and can be adjusted when the workshop is expanded or the equipment is upgraded; Step 1.6: Seamlessly integrate the model with the digital twin system to achieve visualization of VR interaction and real-time monitoring.

3. The method for intelligent HVAC monitoring of Class 1000 cleanrooms based on digital twins according to claim 1 is characterized in that: The system configuration and control layer described in step 2 configures physical devices, IoT sensor networks, PLC controllers, and data transmission modules to achieve real-time collection, processing, and transmission of equipment operating status and environmental parameters, as follows: Step 2.1: During layout, according to the functional requirements of the workshop, through digital twin modeling, use the collaborative relationship matrix between equipment to optimize equipment layout and operation mode; The collaborative relationship matrix M between devices eq (i,j) is: Among them C i,j represents the runtime dependency between devices i and j, M eq (i,j) is used to analyze the collaborative working efficiency of equipment, so as to dynamically adjust the equipment layout and operation priority; Step 2.2: Deploy multiple types of IoT sensor networks at key locations in the workshop, including temperature and humidity sensors, cleanliness sensors, differential pressure sensors, and energy consumption sensors, to conduct real-time monitoring, data integration, and dynamic grouping, and use weighted clustering algorithms for optimization; Step 2.3: The system processes the collected environmental parameters and equipment operation data in real time through the PLC controller, and dynamically adjusts the fan speed, air supply volume and cooling water flow according to the changes in environmental parameters to ensure a stable workshop environment; Step 2.4: The data transmission module uploads the integrated sensor data to the digital twin system and the central platform. Through switches and hierarchical data transmission strategies, the data transmission module achieves the following features: Priority transmission: give priority to abnormal data; Multi-protocol support: The data transmission module supports multiple communication protocols and is compatible with different devices and sensor interfaces; During data upload, a priority queue scheduling algorithm is used to prioritize data packets according to their importance, thus reducing transmission delays for critical data. The calculation formula of queue weight Wdata is as follows: Wdata=α·Turgency+β·Ssize+γ·Rreliability Turgency represents the urgency of data; Ssize is the size of the data packet; Rreliability is the reliability requirement of the data packet; α, β, γ are adjustment coefficients; Step 2.5: When the system detects an abnormal environment, it automatically triggers a response mechanism to dynamically adjust the wind speed, optimize multiple devices, and dynamically adjust the operating status of the equipment; Dynamic wind speed adjustment: When the pressure difference is lower than the set value, the FFU fan speed will automatically increase to maintain a positive pressure environment in the clean area; Multi-device collaborative optimization: The system adjusts the operation mode of the fresh air unit and the exhaust fan unit according to the real-time status of the clean area and the buffer zone, and restores the target environmental parameters at the same time; The system uses a distributed control algorithm to dynamically adjust the operating status of the equipment. When the pressure difference decreases, the system adjusts the fan speed through the following optimization formula: To adjust the front fan speed, To adjust the fan speed: Where ΔP is the target pressure difference P set With the current pressure difference P current The difference between the two, namely ΔP = P set -P current ; K p , K i , K d PID gain parameters after dynamic optimization for reinforcement learning; The system combines predictive control to adjust the operating parameters of other equipment in advance to achieve environmental stability; Step 2.6: To ensure the stability of data collection and equipment operation, configure electrical modules including UPS and power distribution cabinet. Based on the UPS power supply system, introduce a fault prediction model to predict possible faults of the power supply module based on the abnormal fluctuation characteristics of the operating data. The calculation formula of abnormal feature A(t) is: Among them, V set 、V current are the set voltage and the actual voltage respectively; I set ,I current They are set current and actual current respectively; When A(t) exceeds the preset threshold, it switches to the backup power supply in advance to ensure the continuity of data acquisition and transmission.

4. The method for intelligent monitoring of HVAC in a Class 1000 cleanroom based on digital twins according to claim 3 is characterized in that: Real-time monitoring, data integration and dynamic grouping are performed as described in step 2.2, and weighted clustering algorithm is used for optimization, as follows: Real-time monitoring: sensors are used to collect temperature, humidity, cleanliness and pressure difference parameters in real time to ensure that the environment meets the standards of a Class 1000 dust-free workshop; Data integration: Sensor data is intelligently aggregated to filter and integrate redundant data collected from multiple points, thereby improving data transmission efficiency and reducing bandwidth usage; Dynamic grouping: The sensor network is grouped according to the needs of different areas to ensure that data from important areas is processed and responded to first; The weighted clustering algorithm formula is: where w i is the weight of sensor i, d(p i ,p c ) is the distance from sensor i to cluster center (p i ,p c ) distance.

5. The method for intelligent monitoring of HVAC in a Class 1000 clean room based on digital twin according to claim 3 is characterized in that: In step 2.3, the PLC controller integrates an adaptive algorithm to learn and optimize the control strategy, giving priority to meeting the temperature and humidity stability requirements of important areas; The reinforcement learning algorithm is used to optimize the control gain parameters and dynamically learn the optimal control strategy under different scenarios. The formula is: u(t)=Q(s,a)=r(s,a)+γa′maxQ(s′,a′) Where u(t) is the control output; Q(s,a) is the value function of state s and action a; r(s,a) is the immediate reward; γa′max is the discount factor, which indicates the importance of long-term returns; Through reinforcement learning algorithms, the PLC controller can adapt to different working conditions and optimize the wind speed, air supply volume and water flow control of HVAC equipment in real time.

6. The method for intelligent HVAC monitoring of Class 1000 cleanrooms based on digital twins according to claim 1 is characterized in that: The AI ​​algorithm optimization layer described in step 3 combines the multi-objective optimization algorithm, time series prediction algorithm and deep learning model to achieve dynamic optimization of equipment operating status and workshop environmental parameters, as follows: Step 3.1, multi-objective optimization algorithm: Based on the workshop environment control objectives, a multi-objective optimization model is constructed, including temperature and humidity control, cleanliness compliance, and energy consumption minimization optimization; During operation, the environmental data and equipment status parameters of each area are analyzed in real time, and the operating status of the fresh air unit, FFU fan, and refrigeration unit equipment are dynamically adjusted to achieve a balance between environmental requirements and energy efficiency; The multi-objective optimization process constructs the optimization objective function, comprehensively considers the equipment operation efficiency and workshop environment requirements, and realizes the dynamic adjustment of the global optimal solution; The objective function is defined as: min F(x)=αE(x)+β|TT set |+γ|HH set |+δ|CC set | Where E(x) is the total energy consumption of the system; T is the real-time temperature, T set is the target temperature; H is the real-time humidity, H set is the target humidity; C is the real-time cleanliness, C set is the target cleanliness; α, β, γ are weight coefficients used to balance the priorities among energy consumption, temperature and humidity, and cleanliness; The multi-objective optimization model combines the real-time parameters of the environment and finds the optimal solution through a fast genetic algorithm based on constraint optimization; the fitness function f(x) of the genetic algorithm is: Prioritize and retain solutions with higher fitness; Step 3.2, time series prediction algorithm: Use LSTM and Prophet models to model the equipment status and environmental parameters, use the time series prediction algorithm to analyze the equipment's historical operation data and real-time collected status data, predict the equipment's operating trend and future changes in the workshop's environmental parameters, and adjust the equipment's operating mode and environmental parameters by predicting the trends of cooling water flow, fan speed, cleanliness, and temperature and humidity; The time series prediction formula is: y t+1 =f(y t ,h t ) Among them, y t is the environment or equipment state at time t, y t+1 is the environment or equipment status at time t+1; h t is the hidden state at time t, learned by the LSTM model; f represents the prediction model; The Prophet model makes predictions by decomposing the time series into trend, seasonality, and residual components: y(t)=g(t)+s(t)+h(t)+∈ t Among them, g(t) is the long-term trend; s(t) is the periodic change; h(t) is the holiday effect; ∈ t is a random error; Combining the long-term prediction capability of Prophet and the short-term dynamic adaptability of LSTM, a multimodal hybrid prediction model is constructed to achieve prediction of equipment operation trends and environmental status changes; Step 3.3, deep learning model: Based on the graph neural network GNN, a correlation model between devices is established to analyze the interaction and dependency of HVAC equipment during operation. Through comprehensive analysis of multi-dimensional data, potential abnormalities of equipment are identified and possible fault propagation paths are predicted. Based on the analysis results, equipment health status scores are generated to evaluate the operational reliability of equipment in real time. GNN can capture the interaction between device nodes and update node status through adjacency matrix and feature matrix. The node state update formula of GNN is: H (k+1) =σ(AH (k) W (k) ) Among them, H (k) is the k-th layer node feature, H (k+1) is the k+1th layer node feature; A is the adjacency matrix, which represents the connection relationship between devices; W (k) is the weight matrix; σ is the activation function; By analyzing the multi-dimensional data of the equipment, GNN generates a health status score S health : Among them, R i is the failure risk of device i; T i is the total running time of device i; Introducing the graph attention network GAT to perform weighted analysis on key device nodes; Step 3.4, Fault prediction and proactive maintenance: Combine deep learning models to predict equipment failures, generate potential fault propagation paths and impact ranges, and recommend specific proactive maintenance plans based on the health scores and fault risk levels generated by the model, including equipment parts replacement recommendations, operating parameter adjustment strategies, and maintenance priority rankings. Step 3.5, dynamic optimization and real-time feedback: Based on prediction and optimization, the AI ​​algorithm optimization layer adjusts the operating parameters of the equipment in real time, including the air supply volume of the fresh air unit, the wind speed of the FFU fan, and the cooling water temperature of the refrigeration unit, to ensure that the equipment is operating in the optimal state; at the same time, the system feeds back the optimized operating status and adjustment suggestions to the digital twin interface for user reference or manual intervention; Generate energy-saving optimization suggestions based on environmental requirements and equipment operating status f opt for: V f opt =arg minF(V f ) Where V f It is a multi-objective function of the impact of wind speed on the environment.

7. The method for intelligent monitoring of HVAC in a Class 1000 cleanroom based on digital twins according to claim 1 is characterized in that: The digital twin and VR interaction layer described in step 4 develops a three-dimensional visualization interface based on the OpenUSD platform to intuitively present the HVAC equipment and environmental parameters in the workshop in a dynamic and real-time manner, as follows: Step 4.1: The digital twin and VR interaction layer builds a three-dimensional visualization model based on the OpenUSD platform, covering the panoramic display of HVAC equipment and environmental parameters in the workshop, and performs real-time dynamic updates and multi-dimensional visualization; Real-time dynamic update: The digital twin model is integrated with the IoT sensor network to receive and update the equipment operating status and environmental parameters in real time, ensuring that the model always reflects the actual operating status of the workshop; Multi-dimensional visualization: The digital twin model supports dynamic display of multi-dimensional parameters, using graphic elements to intuitively present airflow distribution, cleanliness changes, and temperature and humidity control effects; Step 4.2: Update the equipment status and environmental data collected by the IoT sensor network in real time with the three-dimensional visual operation interface to ensure that users can grasp the environmental changes and equipment operation status in the workshop in real time, so as to achieve management and control. By integrating the IoT network and AI optimization algorithm into the digital twin system, real-time monitoring and predictive analysis functions can be realized: Real-time data flow mapping: The temperature, humidity, cleanliness and pressure difference data collected by IoT sensors are mapped to the three-dimensional digital twin model in real time, and the parameter changes are displayed by dynamic numerical overlay; AI model-driven optimization: Using the prediction results of the AI ​​algorithm, dynamically adjust the display of equipment operating status on the digital twin interface; The real-time mapping formula transforms the sensor data S i (t) is converted into the dynamic display value M of the digital twin model i (t): Among them, S i (t) is the sensor acquisition value at time t; S min , S max It is the minimum and maximum value of the sensor data, used for normalized display; Step 4.3: Combined with virtual reality VR technology, users can realize real-time monitoring, simulated operation and debugging, and virtual fault drill functions in a virtual environment: Real-time monitoring: Users can view the operating status and environmental parameters of the equipment through the VR interface; Simulation operation and debugging: Users adjust equipment operating parameters and observe the impact of these adjustments on the environment in real time, providing intuitive feedback; Virtual fault drill: simulate equipment fault scenarios in a VR environment, analyze fault propagation paths, and generate fault handling suggestions through an interactive visual interface; Using motion capture and user behavior analysis functions, we can record the user's operating behavior in the VR environment in real time, optimize the interface design and interaction efficiency, and evaluate the effectiveness of the user's adjustment parameters. The formula is: in, Parameters adjusted by the user; is the optimal parameter recommended by AI; N is the number of adjustments; The digital twin and VR interaction layer is analyzed by E eff to optimize user experience and improve operational efficiency; Step 4.4: Integrate data from different areas and equipment in the digital twin interface to generate a multi-scenario thermal map of the workshop environment, dynamically displaying the temperature, humidity, cleanliness, and airflow distribution of each area; Step 4.5: The digital twin and VR interaction layer generates targeted optimization suggestions through real-time data analysis and user operation feedback. When areas with insufficient cleanliness are found in the VR environment, the system automatically generates the following optimization solutions: (1) Adjust the fan speed or air supply; (2) Improve the operating efficiency of air-conditioning units; (3) Optimize air flow distribution direction to reduce the risk of pollutant diffusion; The feedback control formula is: u(t+1)=u(t)+K·ΔS Where u(t+1) is the adjusted equipment parameter; ΔS=S set -S current is the deviation between the set value and the current value; K is the feedback gain coefficient, which adjusts the optimization intensity according to the environmental response.

8. The method for intelligent monitoring of HVAC in a Class 1000 cleanroom based on digital twins according to claim 7 is characterized in that: In step 4.4, a multi-scenario heat map of the workshop environment is generated, including: Thermal map generation: The workshop is divided into multiple small units through grid modeling technology, and the parameter value of each unit is calculated in real time through sensor data; Abnormal highlighting: When the parameters of a certain area deviate from the set values, the heat map automatically highlights the abnormal area to help users locate the problem; The formula for generating the heat map is: Where H(x,y,t) is the environmental parameter value at the position (x,y,t) at time t; S i (x, y, t) is the collected value of sensor i at this position; w i is the weight of sensor i, which is set based on distance or data credibility.

9. The method for intelligent monitoring of HVAC in a Class 1000 cleanroom based on digital twins according to claim 1 is characterized in that: The result prediction and feedback layer described in step 5 generates a device health status curve through a time series prediction algorithm, intuitively displays the device health score and trend changes, and enables users to understand the device operation status in real time, as follows: Step 5.1, Equipment health status assessment: The result prediction and feedback layer first conducts a comprehensive analysis of the historical and real-time data of equipment operation based on the time series prediction algorithms LSTM and Prophet, generates an equipment health status curve, analyzes and displays the health score and change trend of the equipment, and allows users to understand the equipment operation status in real time through the status curve, and identify equipment with restored operating efficiency or potential failures in advance; finally, the equipment health status curve is generated and the health score is calculated; Health status curve generation: The system uses historical data and real-time monitoring data to predict the future operating status of the equipment, ensuring that users can intuitively grasp the health change trend of the equipment; Health score model: Calculate the health score of the device based on the deviation between the operating data and the set threshold. The formula is: Among them, M i It is the real-time monitoring value of the equipment operation; M optimal is the optimal operating parameter of the equipment; N is the number of monitoring points in the time window; When S health When the value falls below the set threshold, the system triggers an alarm and recommends proactive maintenance; By combining the short-term dynamic learning ability of LSTM and the long-term trend decomposition model of Prophet, a dynamic balance between short-term warning and long-term maintenance plan is achieved; Step 5.2, Fault propagation path prediction: Use graph neural network (GNN) to establish a dependency model between devices, analyze the possible fault propagation paths of the devices, and predict the impact range of the fault. The system helps users identify the fault propagation path or environmental area by marking the potential impact area of ​​the fault, providing a reference for fault handling; Step 5.3, generate operation optimization suggestions: through multi-objective optimization algorithm, according to the needs of the workshop environment and the operating status of the equipment, dynamically generate optimization suggestions for equipment operation, adjust the wind speed, cooling water flow or new fan unit air supply according to real-time data, ensure that the equipment operates in the best state, and meet the workshop environment control objectives; A multi-objective optimization model is constructed to minimize energy consumption while meeting the temperature, humidity and cleanliness requirements; the objective function is: minE=αE fan +βE cooling +γE ventilation Among them, E fan is the fan energy consumption, which is related to wind speed; E cooling is the energy consumption of the cooling system, which is related to the cooling water flow rate; E ventilation is the energy consumption of the fresh air unit, which is related to the air supply volume; α, β, and γ are weight coefficients used to balance the importance of different energy consumption targets; The constraints are: Temperature constraint: T min ≤T room ≤T max Humidity constraint: H min ≤H room ≤H max Cleanliness constraint: C room ≥C target Pressure difference constraint: ΔP ≥ ΔP target Dynamic feedback control formulas optimize equipment operation modes based on real-time feedback; The improved dynamic feedback control formula is: Among them, V f opt is the optimized wind speed of the fan; e(t) is the deviation between the target value and the actual value; K p , K i , K d is the gain parameter after dynamic optimization; Step 5.4, Operation prediction and energy saving analysis: Combined with optimization suggestions and health status assessment, the operation prediction and assessment module equipment's optimal operating status and potential risks are evaluated, and energy saving solutions are provided based on the current status; Step 5.5, proactive maintenance and feedback plan: Based on the health status curve and fault prediction results, the system generates a maintenance plan, including equipment replacement suggestions and operating parameter adjustment strategies. The maintenance plan is fed back to the user in real time through the digital twin interface, as follows: Step 5.5.1, the health status curve is based on time series data, combined with real-time monitoring values ​​and historical operation data to generate the change trend of the equipment's health score over time, which is used to identify potential faults; The health score calculation formula is: Among them, S health (t) is the health score of the device at time t; x current (t) is the current operating parameter of the equipment; x optimal The optimal operating parameters of the equipment; When S health (t) is below the threshold, the system triggers an early warning and recommends maintenance; Step 5.5.2, time series prediction model: Use the LSTM model to predict future health status curves: Among them, X t Other relevant environmental parameters, including temperature, humidity and cleanliness; Fault prediction generates fault propagation paths and marks the impact range based on the dependencies between devices and historical data; Step 5.5.3, Fault propagation path prediction model: Based on the graph neural network GNN, build a correlation graph between devices: H (l+1) =σ(AH (l) W (l) ) Among them, H (l+1) is the node feature matrix of the lth layer, i.e., the device status; A is the adjacency matrix, which represents the connection relationship between devices; W (l) is the weight matrix; σ is the activation function; Analyze the dependencies between devices and potential impact areas through fault propagation paths: Among them, R fault To spread the risk; i is the path length from the fault source to device i; Generate maintenance plan Generate a specific proactive maintenance plan based on the health status score and fault propagation path, including: Step 5.5.4, equipment replacement recommendation: Identify the equipment that needs maintenance based on the health status score curve and recommend parts replacement. The calculation formula for equipment that needs maintenance is: C replace ={i∣S health (i)<S threshold } Among them, C replace A collection of equipment that needs to be replaced; Step 5.5.5, operating parameter adjustment strategy: adjust the equipment operating parameters, the adjustment formula is: Among them, α is the adjustment step size; is the gradient of equipment energy consumption versus operating parameters; The maintenance plan is fed back in real time through the digital twin interface. Users can view the equipment status and fault areas in a virtual environment and adjust the operating parameters. The real-time feedback formula is: V feedback =φ(S health ,R fault ) Among them, V feedback is the feedback content; φ is the mapping function used to transform the health score and failure risk into suggestions; Feedback includes: fault location and impact range highlighted; The dynamic display maintenance priorities are: P priority =λ1·R fault +λ2·(1-S health ) Step 5.6, real-time data closed-loop update: The system continuously updates the health assessment, fault prediction and operation optimization models through the real-time collected sensor status data, optimizes the later operation data, and adjusts the plan to synchronously feedback to the digital brother life interface to form a closed-loop management mechanism; Combined with real-time data, the system forms a closed-loop control mechanism to continuously optimize operation and maintenance; The closed-loop control formula is: Among them, Δx is the feedback adjustment amount; β is the feedback gain; Finally, the adjustment results are compared with the actual status to generate the next optimization plan.

10. A Class 1000 clean room HVAC intelligent monitoring system based on digital twins, characterized in that: The system is used to implement the intelligent monitoring method for HVAC in a Class 1000 clean room based on digital twins as described in any one of claims 1 to 9, and the system includes a SolidWorks modeling layer, a system configuration and control layer, an AI algorithm optimization layer, a digital twin and VR interaction layer, and a result prediction and feedback layer, wherein: SolidWorks modeling layer: Use SolidWorks software to perform 3D modeling of the HVAC equipment and environment of the Class 1000 clean room; System configuration and control layer: configure physical devices, IoT sensor networks, PLC controllers, and data transmission modules to achieve real-time collection, processing, and transmission of equipment operating status and environmental parameters; AI algorithm optimization layer: Combines multi-objective optimization algorithms, time series prediction algorithms, and deep learning models to achieve dynamic optimization of equipment operating status and workshop environmental parameters; Digital twin and VR interaction layer: Develop a three-dimensional visualization interface based on the OpenUSD platform to intuitively present the HVAC equipment and environmental parameters in the workshop in a dynamic and real-time manner; Result prediction and feedback layer: Generate equipment health status curve through time series prediction algorithm, intuitively display equipment health score and trend changes, so that users can understand equipment operation status in real time.

Images