Building facility full life cycle operation and maintenance monitoring system based on digital twinning

Abstract

The invention relates to the technical field of building facility operation and maintenance, and discloses a building facility full life cycle operation and maintenance monitoring system based on digital twinning. The system comprises a data acquisition module, a digital twin modeling module, an experience library forming module, a state analysis module, a strategy generation module, a strategy adjustment module and an instruction execution module. The data acquisition module acquires real-time and historical operation and maintenance data, and generates standardized data through preprocessing and feature extraction; the digital twin modeling module constructs a model and simulates a running state; the experience library forming module classifies historical operation and maintenance strategies and fault records to form a multi-layer experience library according to strategy confidence and effect indexes. The state analysis module combines the model and the real-time data to evaluate the state and detect abnormity; the strategy generation module generates an operation and maintenance scheme according to the analysis result and an experience library; the strategy adjusting module optimizes the scheme according to the real-time environment and the equipment state; and the instruction execution module converts the optimization strategy into a control instruction and sends the control instruction to execution equipment to adapt to the full-life-cycle operation and maintenance requirements of the building facilities.

Description

Technical Field

[0001] This invention relates to the field of building facility operation and maintenance technology, specifically a building facility full life cycle operation and maintenance monitoring system based on digital twins. Background Technology

[0002] As the construction industry continues to develop, the scale and complexity of building facilities are constantly increasing, ranging from ordinary civil buildings to large commercial complexes and industrial plants. The equipment systems they contain are becoming increasingly diverse, encompassing multiple categories such as HVAC, water supply and drainage, electrical equipment, and fire protection systems. The coordinated operation of these devices directly affects the overall performance and safety of the building facilities. With the popularization of the concept of full life-cycle management of building facilities, the operation and maintenance phase, as a crucial link in the use of building facilities, has received widespread attention in the industry regarding its management quality and efficiency.

[0003] Currently, the operation and maintenance management of building facilities mainly relies on traditional manual inspections and experience-based judgment. In terms of data collection, manual inspections are not only inefficient but also susceptible to human factors, leading to data lag, incompleteness, and inaccuracy, making it difficult to reflect the actual operating status of building facilities and equipment in real time. In scenarios using automated monitoring equipment, differences in data formats and collection frequencies among different devices, coupled with a lack of unified data processing standards, prevent the effective integration and sharing of monitoring data, creating "data silos" that hinder the realization of data value.

[0004] At the model building and status analysis level, traditional operation and maintenance management lacks the ability to digitally model building facilities, making it impossible to construct virtual models that are highly consistent with physical facilities. This hinders the dynamic simulation and visualization of facility operating status. During status assessment and anomaly detection, operation and maintenance personnel often rely on personal experience for judgment. For complex equipment systems, potential faults are difficult to identify and warn of in advance, often requiring repairs only after a fault occurs. This not only increases maintenance costs but may also lead to facility downtime, affecting normal building use and even causing safety accidents.

[0005] In terms of operation and maintenance strategy formulation and adjustment, traditional models often rely on fixed maintenance cycles or past experience, lacking the ability to dynamically respond to real-time operational data and environmental changes. When building facilities face changes in the external environment (such as temperature and humidity fluctuations), differences in equipment aging, or changes in usage load, fixed operation and maintenance strategies are difficult to adapt to actual needs, potentially leading to over-maintenance or under-maintenance. Over-maintenance wastes human and material resources, while under-maintenance increases the risk of equipment failure. Furthermore, the industry lacks a systematic review and classification of historical operation and maintenance strategies and fault handling experience, making it difficult to effectively reuse past experience. This results in the need to re-explore solutions when similar problems recur, reducing the efficiency of operation and maintenance management. Summary of the Invention

[0006] The purpose of this invention is to provide a digital twin-based full lifecycle operation and maintenance monitoring system for building facilities to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides a building facility full lifecycle operation and maintenance monitoring system based on digital twins, the system comprising:

[0008] The data acquisition module acquires real-time monitoring data and historical operation and maintenance data of building facilities, preprocesses and extracts features from the data, and generates standardized data.

[0009] The digital twin modeling module constructs a digital twin model of the building facility based on the standardized data and simulates its operating status.

[0010] The experience base formation module classifies historical operation and maintenance strategies and fault handling records based on strategy confidence and effectiveness indicators to form a multi-layered experience base.

[0011] The status analysis module performs status assessment and anomaly detection based on the digital twin model and real-time monitoring data, and outputs status analysis results.

[0012] The strategy generation module generates an operation and maintenance strategy plan based on the status analysis results and the multi-layer experience base.

[0013] The strategy adjustment module dynamically adjusts the operation and maintenance strategy based on real-time environmental changes and equipment status, and outputs optimized control strategies.

[0014] The instruction execution module generates control instructions based on the optimized control strategy and sends them to the execution device.

[0015] Preferably, the standardized data includes data feature vectors, time series data sets, and data quality indicators; the digital twin model includes a three-dimensional geometric model, a physical attribute model, and a behavioral simulation model; the multi-layered experience base includes strategy confidence distribution, effect indicator ranking, and historical strategy set; the status analysis results include status scores, anomaly warning indicators, and predicted trend data; the operation and maintenance strategy scheme includes maintenance plans, parameter adjustment, and execution steps; the optimization control strategy includes real-time instructions, parameter adjustment, and execution priority; the control instructions include equipment control commands and parameter settings; and the feedback data includes execution results and equipment status changes.

[0016] Preferably, the data acquisition module includes:

[0017] The data collection submodule acquires real-time sensor data and historical operation and maintenance data of building facilities, and records the data source and timestamp;

[0018] The data preprocessing submodule performs data cleaning and format conversion based on the data output by the data collection submodule, generating cleaned data;

[0019] The feature extraction submodule extracts data feature vectors and time series features from the cleaned data output by the data preprocessing submodule to generate standardized data.

[0020] Preferably, the digital twin modeling module includes:

[0021] The model building submodule constructs a three-dimensional geometric model and a physical property model of the building facilities based on the standardized data;

[0022] The behavior simulation submodule constructs a model based on the model output by the submodule, simulates the operational behavior of building facilities, and generates a digital twin model.

[0023] Preferably, the experience base formation module includes:

[0024] The strategy collection submodule retrieves historical operation and maintenance strategies and fault handling records;

[0025] The effect evaluation submodule collects the data output by the submodule based on the strategy, and evaluates the strategy execution effect and confidence level;

[0026] The classification and storage submodule classifies and stores the strategies according to the evaluation results output by the effect evaluation submodule, forming a multi-layered experience base.

[0027] Preferably, the state analysis module includes:

[0028] The data input submodule acquires the digital twin model and real-time monitoring data;

[0029] The status calculation submodule calculates the status score and health level based on the data output by the data input submodule.

[0030] The anomaly detection submodule performs anomaly detection and generates early warnings based on the results output by the state calculation submodule.

[0031] The output module integrates status and exception information, and outputs status analysis results.

[0032] Preferably, the strategy generation module includes:

[0033] The strategy query submodule queries similar strategies from a multi-layered experience base based on the state analysis results;

[0034] The strategy optimization submodule optimizes the strategy based on the queried strategy and the current status, and generates an operation and maintenance strategy plan.

[0035] Preferably, the strategy adjustment module includes:

[0036] The real-time data acquisition submodule obtains real-time environmental changes and equipment status;

[0037] The dynamic adjustment submodule adjusts the operation and maintenance strategy based on the data output by the real-time data acquisition submodule, and outputs optimized control strategies.

[0038] Preferably, the instruction execution module includes:

[0039] The instruction generation submodule generates equipment control instructions based on the optimized control strategy;

[0040] The instruction sending submodule sends control instructions to the execution device.

[0041] Preferably, the system further includes: a feedback update module, used to collect feedback data after execution and update the digital twin model and the multi-layer experience base according to the feedback data;

[0042] The feedback update module includes:

[0043] The feedback collection submodule collects the device status and data changes after execution;

[0044] The model update submodule collects the data output by the submodule based on feedback and updates the digital twin model;

[0045] The experience base update submodule updates the multi-level experience base based on feedback data and execution results.

[0046] Compared with the prior art, the beneficial effects of the present invention are:

[0047] In terms of data processing, the data acquisition module can simultaneously acquire real-time monitoring data and historical operation and maintenance data of building facilities, and generate standardized data through preprocessing and feature extraction. This process effectively solves problems such as lagging data acquisition, inconsistent formats, and "data silos" in traditional operation and maintenance, realizing the integration and standardization of data from different sources and in different formats, ensuring the integrity, accuracy, and consistency of the data, and providing a reliable data foundation for subsequent model building, status analysis, and other stages. Standardized data not only facilitates efficient flow and sharing between various modules within the system, but also breaks down data barriers between different devices and systems, fully releasing the value of data, enabling operation and maintenance personnel to carry out their work based on comprehensive and accurate data, and avoiding decision-making biases caused by data problems.

[0048] At the digital twin modeling level, the digital twin modeling module constructs digital twin models of building facilities based on standardized data and simulates their operational status. This model maintains a high degree of consistency with the physical building facilities, enabling dynamic simulation and visualization of the facility's operational status. Maintenance personnel can intuitively understand the real-time operational status of building facilities and equipment through the virtual model, grasping the overall condition of the facilities without relying on manual inspections. This visualization and dynamic simulation capability not only reduces the workload of maintenance personnel but also helps them grasp the operational patterns of the facilities more quickly and accurately, promptly identifying potential operational anomalies and overcoming the limitations of traditional maintenance methods that rely on manual observation and experience-based judgment.

[0049] The experience base generation module organizes historical operation and maintenance strategies and fault handling records, classifying them based on strategy confidence and effectiveness metrics to form a multi-layered experience base. This design transforms previously scattered and disorganized operation and maintenance experience into structured and reusable knowledge resources, avoiding the problem of difficulty in inheriting and reusing experience in traditional operation and maintenance. When encountering similar facility status or fault issues, the strategy generation module can directly retrieve relevant experience from the multi-layered experience base without having to explore solutions again, significantly shortening strategy formulation time and improving the efficiency of operation and maintenance decision-making. At the same time, the multi-layered classification method ensures the orderliness and relevance of the experience base; strategies with different confidence levels and effectiveness metrics can correspond to different facility operation scenarios, making strategy selection more aligned with actual needs.

[0050] The status analysis module combines a digital twin model with real-time monitoring data to conduct status assessments and anomaly detection. Compared to traditional operation and maintenance methods that rely on human experience, this module can leverage the dynamic simulation capabilities of the model and the timeliness of real-time data to more accurately assess the facility's operational status. By comparing and analyzing real-time data with model simulation data, abnormal situations in facility operation can be detected in a timely manner, including potential fault hazards. This enables proactive and accurate anomaly detection, changing the passive "reactive maintenance" approach of traditional operation and maintenance. It can effectively reduce the probability of failures and lower losses and maintenance costs caused by facility downtime.

[0051] The strategy generation module generates operation and maintenance strategy solutions based on status analysis results and a multi-layered experience base, ensuring the scientific nature and relevance of the strategies. This module does not rely on fixed experience or periodic strategy development; instead, it combines real-time facility status with categorized and validated historical experience. This allows the generated strategies to accurately match current facility operational needs, avoiding the problems of over-maintenance or under-maintenance in traditional operations and maintenance. While ensuring normal facility operation, it rationally allocates operation and maintenance resources, reducing resource waste.

[0052] The strategy adjustment module dynamically adjusts the operation and maintenance strategy based on real-time environmental changes and equipment status, further enhancing the adaptability of the strategy. During the operation of building facilities, factors such as the external environment (e.g., temperature, humidity), equipment load, and equipment aging may change. Fixed operation and maintenance strategies are difficult to cope with these dynamic changes, but this module can sense these changes in real time and optimize and adjust the strategy accordingly. This ensures that the operation and maintenance strategy is always consistent with the actual operating conditions of the facility, effectively responding to various emergencies or environmental fluctuations, and guaranteeing the stability and reliability of facility operation.

[0053] The instruction execution module generates control instructions based on the optimized control strategy and sends them to the execution equipment, enabling rapid implementation and closed-loop management of the operation and maintenance strategy. This module transforms the strategy into specific, executable instructions, ensuring that the execution equipment can respond promptly to strategy adjustments. This avoids the problem of strategy and execution being disconnected in traditional operation and maintenance, forming a complete closed loop from data collection, status analysis, strategy formulation, strategy adjustment to instruction execution. This improves the overall efficiency and collaboration of operation and maintenance management, promotes the development of building facility operation and maintenance management towards intelligence and refinement, and meets the needs of modern building facility full lifecycle operation and maintenance management. Attached Figure Description

[0054] Figure 1 This is a timing diagram of the digital twin-based building facility lifecycle operation and maintenance monitoring system described in this invention.

[0055] Figure 2 A flowchart illustrating the operation of the status analysis module. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Please see Figure 1 This invention provides a building facility full lifecycle operation and maintenance monitoring system based on digital twins. The system includes: a data acquisition module, a digital twin modeling module, an experience base formation module, a status analysis module, a strategy generation module, a strategy adjustment module, and an instruction execution module. Specific implementation methods are as follows:

[0058] The data acquisition module acquires real-time monitoring data and historical operation and maintenance data of building facilities, preprocesses and extracts features from the data to generate standardized data. The digital twin modeling module constructs a digital twin model of the building facilities based on the standardized data and simulates its operating status. The experience base formation module classifies historical operation and maintenance strategies and fault handling records based on strategy confidence and effectiveness indicators to form a multi-layered experience base. The status analysis module performs status assessment and anomaly detection based on the digital twin model and real-time monitoring data, and outputs status analysis results. The strategy generation module generates operation and maintenance strategy plans based on the status analysis results and the multi-layered experience base. The strategy adjustment module dynamically adjusts the operation and maintenance strategy plans based on real-time environmental changes and equipment status, and outputs optimized control strategies. The instruction execution module generates control instructions based on the optimized control strategies and sends them to the execution equipment. Through the coordinated operation of these modules, the system achieves intelligent and automated management of building facility operation and maintenance.

[0059] Example 1: The generation process of standardized data involves multi-dimensional data processing. Data feature vectors extract key attributes from the original monitoring data using feature engineering methods. For example, for air conditioning system operation data, feature values ​​such as compressor frequency, condensing temperature, and supply air temperature difference are extracted to form a structured feature array. The time series dataset uses a time series database to store continuous monitoring data such as temperature, humidity, and vibration. Each data point is associated with a timestamp accurate to the millisecond level and organized into fixed-length data segments using a sliding window mechanism. Data quality indicators include three dimensions: completeness rate, outlier ratio, and timeliness. The completeness rate is calculated as the percentage of valid data points out of the total sampling points. The outlier ratio is identified using a box plot method. Timeliness records the delay time from data acquisition to processing completion. Data feature vectors are stored in a feature library after dimensionality reduction through principal component analysis. The time series dataset is partitioned and stored by device number and a time index is established. Data quality indicators generate metadata files in JSON format.

[0060] The digital twin model's 3D geometric model is generated based on the BIM model conversion. It uses IFC format to analyze the coordinates of building structure beams and columns, equipment spatial locations, and pipeline topology, employing lightweight processing to retain key geometric features. The physical property model integrates equipment nameplate parameters and material property databases. HVAC equipment includes dynamic parameters such as COP value and rated power, while the building envelope stores static parameters such as thermal conductivity and heat capacity. The behavioral simulation model uses multiphysics coupling simulation. Structural behavior simulation calculates load deformation through finite element analysis, fluid behavior uses CFD to simulate wind field distribution, and electrical behavior calculates energy consumption fluctuations through circuit simulation. The model update mechanism uses incremental updates. When sensors detect equipment replacement, the physical property model automatically synchronizes with the new equipment parameters. When the deviation between monitoring data and simulation results exceeds 5%, behavioral model calibration is triggered.

[0061] The multi-layered experience base is constructed using a hierarchical storage architecture. Policy confidence distribution is modeled using a Bayesian probability model based on device type to calculate the posterior probability of successful policy execution under different operating conditions. The TOPSIS multi-criteria decision-making method is used to rank performance indicators, generating a policy priority list by comprehensively considering six dimensions, including fault repair time, resource consumption, and system impact. The historical policy set is stored in a graph database, with nodes representing device fault types and edges storing associated processing solutions and execution records. Experience base retrieval employs a hybrid indexing mechanism, using a hash index based on fault codes to quickly locate basic policies and a range index based on status scores to filter applicable solutions. The confidence update module automatically adjusts the probability distribution based on actual results after each policy execution. Performance indicators are re-weighted and recalculated quarterly, with ranking weights adjusted according to the latest operational goals.

[0062] The calculation process for the status analysis results includes multi-source data fusion. The status score adopts the fuzzy comprehensive evaluation method, integrating equipment operating parameters, environmental data, and historical fault records to output a health index of 0-100 points. Anomaly warning indicators use a three-level classification mechanism: Level 1 warnings are for parameters exceeding limits (e.g., temperature exceeding a threshold); Level 2 warnings are for abnormal trends (e.g., continuously increasing vibration acceleration); and Level 3 warnings are for complex fault modes (e.g., sudden drop in energy efficiency accompanied by abnormal noise). Predictive trend data is generated using an LSTM time-series prediction model, outputting key parameter change curves for the next 72 hours, with the prediction confidence interval updated every 15 minutes. The analysis results visualization module generates a heatmap to mark the equipment health status, using red, yellow, and green to mark abnormal areas in the 3D model, and trend data is overlaid on the equipment model as dynamic charts.

[0063] The generation logic of the operation and maintenance strategy scheme follows a case-based reasoning mechanism. Maintenance plans automatically schedule preventative maintenance time windows to avoid peak equipment operating hours. The parameter adjustment list includes adjustable variables such as inverter frequency setpoints and valve opening corrections. Execution steps are decomposed into atomic operation commands, such as "close VAV-203 air valve → start standby pump P-7 → adjust frequency to 45Hz". The scheme verification module pre-simulates the strategy execution process in a digital twin model, detecting potential equipment conflicts or system oscillation risks and outputting a safety verification report. Dynamic adjustment of the optimized control strategy employs a reinforcement learning framework. Real-time commands scan environmental parameter changes every 5 seconds, automatically lowering the chilled water setpoint when outdoor temperature rises sharply. Parameter settings utilize adaptive dead-zone control to avoid frequent actuator actions. Execution priority uses a preemptive scheduling algorithm, with safety-related commands (such as fire alarm linkage) set to highest priority and energy efficiency optimization commands set to medium priority. The strategy optimization engine continuously monitors KPI indicators, automatically triggering a strategy regeneration process when the overall system energy efficiency ratio decreases by 1.5%. Control commands are generated according to industrial communication protocol standards. Equipment control commands are encapsulated in Modbus RTU format for start / stop instructions, and HVAC equipment includes mode switching codes (cooling / heating / dehumidification). Parameter setting values ​​are transmitted as floating-point values ​​using the OPCUA protocol, retaining two decimal places. A command safety verification module verifies that command values ​​are within the equipment's allowable range to prevent damage from over-limit operations. Feedback data acquisition uses a dual-channel mechanism; execution results are returned as operation status codes (success / failure / timeout) by the equipment controller; equipment status changes are captured in real-time by a sensor network, comparing parameter change rates before and after command execution. The data synchronization module starts a 15-second countdown after a command is issued; if no feedback is received within the timeout period, the command is marked as abnormal, triggering a fault diagnosis process. The feedback analysis unit calculates the strategy execution deviation; when the actual temperature adjustment does not meet expectations, it records the deviation value and environmental interference factors.

[0064] Example 2: The data acquisition module's operation begins with the sensor network scan of the data collection submodule. This submodule connects to sensing devices distributed throughout the building, such as temperature and humidity sensors, current transformers, and vibration monitors, via an industrial IoT gateway. It polls and collects real-time operating parameters every 30 seconds, while simultaneously extracting maintenance work order records from the building equipment management system's historical database for the past five years. All collected data is appended with device ID codes and a timestamp accurate to milliseconds, forming a raw data stream with metadata. The data preprocessing submodule uses a pipelined architecture to process the raw data stream. The data cleaning unit first identifies and removes outliers that significantly exceed the physical range (such as indoor temperature readings of -20℃), fills missing data using the mean interpolation method, and removes duplicate data based on timestamps. The format conversion unit converts heterogeneous data into a unified JSON format, retaining one decimal place for temperature data, converting current values ​​to amperes, and preserving the original accelerometer voltage values ​​for vibration data. The feature extraction submodule performs multi-level feature mining on the preprocessed data. The data feature vector generation unit calculates statistical features (such as the mean, variance, and kurtosis of the air conditioner return air temperature within 10 minutes) and extracts frequency domain features (such as the main frequency amplitude of the FFT spectrum of the water pump vibration signal). The time series feature unit uses the sliding window technique to calculate the trend slope, periodic components, and fluctuation entropy values ​​of the power load data within a 60-minute window. The final output standardized data package includes a feature vector array, a time series segment matrix, and a data quality report recording the cleaning process.

[0065] The model building submodule of the digital twin modeling module adopts parametric modeling technology. The 3D geometric model unit imports Building Information Modeling (BIM) files, retaining key structural features through geometric topology simplification algorithms, and uses a global building positioning system for the spatial coordinate system. The physical property model unit is linked to the equipment and material database; for water supply and drainage systems, pipe diameter, material, and roughness coefficient are labeled; for electrical distribution cabinet components, electrical parameters such as resistance and insulation class are entered; and for HVAC ducts, the thermal conductivity of the insulation layer is defined. The behavior simulation submodule constructs a multiphysics coupling engine. The structural behavior simulation unit performs static finite element analysis on the foundation of large equipment, calculating stress distribution under different load conditions; the fluid behavior unit uses computational fluid dynamics to simulate temperature stratification in room air conditioning; and the electrical behavior unit establishes an equivalent circuit model to predict voltage fluctuations in the power distribution system during sudden load changes. The model verification mechanism compares sensor measured data with simulation results in real time. When the deviation between the simulated chilled water flow rate and the ultrasonic flow meter reading exceeds 3%, a pipe resistance coefficient calibration procedure is automatically triggered.

[0066] The collaboration between the data acquisition module and the digital twin modeling module is achieved through a data pipeline. Standardized data output from the feature extraction submodule is directly input into the model construction submodule. When a newly installed variable air volume (VAV) terminal device is identified by the sensor network, the model construction submodule automatically adds the device geometry to the 3D scene and simultaneously imports the rated air volume parameters from the device manual. The behavior simulation submodule initiates a full-system dynamic simulation every morning to predict the energy consumption distribution for the next 24 hours, and the simulation results are stored as a behavior baseline curve for the status analysis module to access. The data anomaly handling mechanism has a three-level response. When the data preprocessing submodule detects that a sensor has failed to acquire data 10 times consecutively, it automatically marks the device as faulty and highlights it in the digital twin model. When the behavior simulation submodule detects a continuous deviation between the simulation results and the measured data, it sends a model calibration prompt to the maintenance personnel.

[0067] Real-time data stream processing employs an edge computing architecture. Intelligent gateways deployed at the device layer perform initial data cleaning, uploading only feature vectors and key time series data to the cloud. The hierarchical loading mechanism of the digital twin model allows for the use of models of varying precision based on analytical needs; simplified physical models are used for routine monitoring, while detailed multiphysics coupling models are loaded for fault diagnosis. Time synchronization services ensure strict alignment of sensor data timestamps with the simulation clock, automatically activating a local clock compensation algorithm when network latency exceeds 500 milliseconds. Model version management uses blockchain technology to record each model update. When a new model of water pump is installed, the physical property model update record includes the old model's retirement time, the new model's technical parameters, and the updating operator's information. The version iteration log of the behavioral simulation model records detailed algorithm optimizations, such as upgrading the wind turbine energy consumption simulation from empirical formulas to a fluid dynamics model based on impeller theory. Lightweight rendering of the 3D geometric model uses LOD technology; a simplified shell model is displayed for remote viewing, while a detailed model including bolt-level precision is loaded during maintenance operations.

[0068] Taking the example of a commercial complex's central air conditioning system experiencing a decline in cooling efficiency during the summer cooling season, the data acquisition module's data collection submodule collects operating parameters every 30 seconds through 128 sensors installed on equipment such as the chiller, water pumps, and cooling towers. These parameters include real-time data such as chiller power (current value 356kW), chilled water supply and return temperatures (7.2℃ / 11.8℃), cooling water inlet and outlet temperatures (32.5℃ / 37.8℃), and water pump flow rate (285m³ / h). Simultaneously, it exports historical operating records from the building automation system for the past three cooling seasons. All data is appended with equipment codes and precise timestamps. The data preprocessing submodule initiates a data cleaning process, identifying and removing obvious outliers (such as an abnormal reading of -5.6℃ instantaneously collected by a temperature sensor). Missing data caused by communication interruptions in some sensors is supplemented using adjacent time-point interpolation. Data collected using different protocols is uniformly converted to standard JSON format, current values ​​are converted to amperes, and temperature values ​​are retained to one decimal place. The feature extraction submodule performs in-depth analysis on the preprocessed data and generates a data feature vector containing key performance indicators of the refrigeration system: real-time coefficient of performance (COP) value of 3.15, total system cooling load rate of 82.7%, and equipment operation stability index of 0.87. The time series feature extraction unit analyzes the chilled water return temperature curve and calculates its rising slope of 0.12℃ / min and fluctuation amplitude of 0.8℃ within 10 minutes.

[0069] The model building submodule of the digital twin modeling module reconstructs the system's 3D model based on standardized data. It extracts the refrigeration unit's geometric dimensions and piping topology from the BIM system to establish a 3D visualization model including major components such as the compressor, condenser, and evaporator. The physical property model unit inputs actual equipment parameters: refrigerant type R134a, compressor rated power 400kW, evaporator heat exchange area 285m², and other physical characteristics. The behavior simulation submodule initiates multiphysics simulation. The fluid dynamics unit simulates the flow velocity distribution and pressure drop characteristics of chilled water in the pipes, the thermodynamics unit calculates the refrigerant's phase change process and energy transfer efficiency in the system, and the structural mechanics unit analyzes the vibration propagation law during pump operation. Simulation results show that under current operating conditions, insufficient cooling water flow leads to a higher condensing temperature, and the system COP value is 12.3% lower than the design value. The model verification mechanism compares simulated data with measured data. The simulated cooling water outlet temperature of 37.6℃ and the measured value of 37.8℃ are within the allowable range, but the simulated chilled water return temperature of 11.2℃ differs significantly from the measured value of 11.8℃. The system automatically triggers a model calibration program to adjust the evaporator heat transfer coefficient model parameters, making the simulated value closer to the measured data. A bidirectional data channel is established between the data acquisition module and the digital twin modeling module. The actual COP value of the system calculated by the feature extraction submodule is transmitted to the digital twin model in real time, and the behavior simulation submodule readjusts its simulation strategy based on this data. Edge computing nodes perform local preprocessing of high-frequency vibration data, uploading only the feature value (vibration acceleration RMS value 2.8mm / s²) to the central system, reducing network data transmission volume. The time synchronization service calibrates the clocks of all sensors, ensuring that the chilled water temperature acquisition time deviates from the simulation clock by no more than 50 milliseconds. When a temperature sensor times out multiple times consecutively, the digital twin model automatically replaces it with simulated data and marks the sensor's fault status in the 3D model.

[0070] The model version management system recorded the entire model calibration process: the initial model version V3.2.1 was calibrated starting at 14:30 on August 2nd, adjusting the evaporator fouling factor from 0.00012 to 0.00018, improving model accuracy by 23%; the calibrated model version V3.2.2 is marked as the current valid version. Lightweight 3D model processing compressed the original 2.3GB detailed model into a 185MB rendered model, preserving key geometric features while improving visualization smoothness. The system automatically runs a full-system simulation every 6 hours to predict energy consumption distribution over the next 24 hours, providing data support for operational decisions.

[0071] Example 3: See Figure 2The strategy collection submodule of the experience base module extracts historical work order data from the computerized maintenance management system via API interface. This data includes structured fields such as fault description text, corrective actions taken, spare parts lists, and personnel information. It also extracts equipment control parameter adjustment records from the SCADA system logs. All strategy data is uniformly coded according to the ISO14224 standard; a water pump fault is marked as "EQP-PUMP-001," and an air conditioning compressor overload is marked as "HVC-COMP-002," forming a standardized set of strategy events. The effectiveness evaluation submodule adopts a multi-dimensional evaluation system, and the strategy execution effect is calculated using comprehensive performance indicators.

[0072]

[0073] Where: R represents the fault repair rate (range 0-1), C represents the resource consumption cost (unit: yuan), D represents the downtime (unit: hours), and α, β, and γ are the weighting coefficients of each indicator (default values ​​0.5, 0.3, 0.2); confidence calculation is based on the strategy execution history, using the ratio of successful executions to total executions combined with data integrity factors for correction. The classification storage submodule adopts an ontology-based knowledge representation method, constructing a three-dimensional classification matrix of strategies according to three dimensions: device type, fault mode, and processing method. Each strategy entry stores the original record, effect indicator, confidence score, and the timestamp of the last successful execution; the storage structure is implemented using a graph database, where nodes represent strategy entities and edges represent the relationships between strategies.

[0074] The data input submodule of the status analysis module establishes a bidirectional data channel, receiving real-time simulation data from the digital twin modeling module and simultaneously acquiring measured data from physical sensors from the data acquisition module. The data alignment unit uses a time-series matching algorithm to synchronize simulation and measured data on a unified time axis, and linear interpolation is used to handle sampling frequency differences. The status calculation submodule implements multi-source information fusion. The status score calculation uses the analytic hierarchy process (AHP) to construct a judgment matrix containing 18 evaluation indicators, including equipment operating parameters, environmental conditions, and historical maintenance records. The weight of each indicator is calculated using the eigenvector method. The health assessment introduces an equipment degradation model, calculating the remaining lifespan percentage based on cumulative load over operating time and wear coefficients of key components. The anomaly detection submodule adopts a hybrid detection strategy. For parameters with clear thresholds, a rule engine triggers a first-level warning. For complex patterns, a deep learning model is used for anomaly pattern recognition, with training data derived from vibration spectrum diagrams of historical failure cases. The warning generation unit classifies anomalies into three levels: attention, warning, and danger, based on severity. Each warning includes the location information of the abnormal equipment, a list of possible causes, and suggested inspection items. The results output submodule generates a structured analysis report, which includes a current status score line chart, a health radar chart, an anomaly warning list, and a key parameter trend prediction chart. The data format uses a standardized JSON schema for easy parsing by subsequent modules.

[0075] The collaboration between the experience base and the status analysis module is achieved through a strategy matching engine. When the anomaly detection submodule identifies excessive vibration in a centrifugal chiller unit, it automatically queries the experience base for vibration-related handling strategies. Query conditions include multi-dimensional filtering conditions such as equipment type "centrifugal chiller," fault symptom "excessive vibration," and current load rate "70%-85%." The weighting coefficients of the effect evaluation submodule can be dynamically adjusted according to the operation and maintenance strategy. During peak electricity consumption seasons, the weighting coefficient β of energy consumption indicators is increased, and during important meeting support periods, the weighting coefficient α of repair rate is increased. The health model of the status calculation submodule supports personalized equipment configuration. For critical refrigeration units, a cumulative damage model based on actual compression operating time is used, while for ordinary fan coil units, a simplified linear operating time model is used. Data timeliness management adopts a sliding window mechanism; status scoring calculations only use operating data from the most recent 30 days, with earlier historical data archived in the long-term trend analysis database. The online learning function of the anomaly detection model is updated monthly, adding newly occurring fault cases to the training dataset. The experience base's version control records every strategy addition and modification. When a strategy's performance falls below a threshold after three consecutive executions, the system automatically marks it as requiring manual review. The status analysis module's verification mechanism compares the digital twin model's predicted values ​​with actual measurements daily. When significant deviations persist, a model calibration process is triggered. The visualization of output results employs adaptive presentation technology, displaying simplified health indices and emergency warnings on mobile devices, and complete multi-dimensional analysis charts on workstations. All status analysis results are linked to a 3D geographic information model; clicking on a warning message automatically locates the device's specific position within the building. Historical status data compression and storage utilizes a rotating door algorithm, retaining key trend characteristics while reducing data storage by 70%. Long-term historical data allows for backtracking and querying status records for any period within the past five years.

[0076] Example 4: The strategy query submodule of the strategy generation module adopts a multi-condition matching algorithm. When the status analysis module outputs a warning of reduced chiller efficiency, the query conditions include multiple dimensions such as equipment type "centrifugal chiller", fault phenomenon "cooling efficiency less than 85%", load rate "70%-80%" and ambient temperature "28-32℃". The query engine first searches for complete matches in the multi-layer experience base. If there are no complete matches, it gradually broadens the range of conditions and finally returns a set of strategies with a similarity higher than 0.8. The strategy optimization submodule is based on a multi-objective optimization framework and adaptively adjusts the candidate strategies found in the query. For example, in the case of efficiency optimization of a data center chiller unit, the original strategy suggestion was "clean the condenser piping + adjust the cooling water flow rate". The optimization module combined the current water hardness data (measured value 185mg / L) and the unit's operating time (has been running continuously for 8 months) to upgrade the cleaning plan from ordinary chemical cleaning to dual-circulation acid washing, while adjusting the cooling water flow rate from the rated value of 85% to 88%. The optimized strategy plan includes specific operating steps, a list of required materials, expected downtime, and safety precautions.

[0077] The real-time data acquisition submodule of the strategy adjustment module continuously acquires field data via the OPCUA protocol, including 32 monitoring parameters such as cooling tower outlet water temperature (real-time value 29.3℃), condenser approach temperature (real-time value 4.2℃), and compressor current (real-time value 285A). The environmental monitoring unit simultaneously collects external variables such as outdoor temperature and humidity (31.5℃ / 65%RH) and atmospheric pressure (100.3kPa). The dynamic adjustment submodule adopts a combination of rule engine and machine learning. When the real-time data shows that the condenser approach temperature suddenly rises to 5.1℃, the strategy adjustment process is automatically triggered: first, the currently executing chemical cleaning operation is paused, and then the cooling water flow setpoint is recalculated based on the real-time load data (current cooling load 72%), adjusting the flow rate from 88% in the original plan to 91%, while adding a condenser fan speed increase command. The adjusted optimized control strategy includes a revised parameter setting sequence and execution priority sorting, with emergency operation commands set to the highest priority for immediate execution, and parameter optimization commands set to the medium priority for planned execution.

[0078] The data recording during the strategy execution process adopts an event-driven log. See Table 1 for the changes in key parameters during the execution of a chiller unit optimization strategy.

[0079] Table 1: Record of Execution Parameters for Chiller Unit Optimization Strategy

[0080] Timestamp Strategy Phase Cooling water flow rate (%) Condensation temperature (°C) Compressor current (A) Cooling efficiency (%) Execution status 08:00:00 initial state 85.0 35.2 298 82.3 - 08:02:15 Traffic Adjustment 88.0 34.8 291 83.1 In progress 08:05:40 Cleaning begins 88.0 34.6 289 83.5 In progress 08:15:30 Mid-term testing 88.0 33.9 285 85.2 Under monitoring 08:25:10 Dynamic adjustment 91.0 33.5 282 86.7 Under adjustment 08:35:00 Completed status 91.0 33.1 280 87.9 Completed

[0081] The coordination between strategy generation and adjustment is achieved through closed-loop control. When the dynamic adjustment submodule detects an abnormal increase in condenser temperature, it not only adjusts the current strategy parameters but also feeds back environmental change information to the strategy generation module. The strategy optimization submodule updates the strategy knowledge base accordingly, recording the adjustment rule of "increasing cooling water flow by 3-5% under high temperature and humidity conditions." The strategy query mechanism supports fuzzy matching. When encountering new fault modes, the system automatically decomposes fault characteristics, queries the handling strategies for abnormal current and temperature separately, and then performs combined optimization. Real-time data acquisition adopts a distributed architecture, deploying edge computing nodes at the equipment site for data preprocessing, uploading only feature values ​​and time series summaries to the central system. The dynamic adjustment submodule retains basic adjustment rules at the edge nodes, enabling the execution of key parameter adjustments even during network interruptions. Strategy version management records the decision basis for each adjustment. When adjusting the cooling water flow from 88% to 91%, the system records the adjustment reason, decision basis, and adjustment effect. The strategy simulation verification function pre-runs the adjustment scheme in a digital twin model, detecting any system conflicts and ensuring the safety of the adjustment strategy. The anomaly handling mechanism includes a multi-level rollback strategy. If the equipment status does not improve after the adjusted parameters are executed, the system automatically rolls back to the previous stable parameter setting and triggers an expert intervention request. All strategy adjustment records form a decision trajectory, including a complete information chain such as timestamp, parameters before adjustment, parameters after adjustment, decision reasons, and execution effects. The strategy optimization knowledge base adopts an incremental learning mechanism, where the experience of each successful adjustment is abstracted into rule conditions, continuously enriching the system's adaptive adjustment capabilities. Cross-system integration is achieved through standard interfaces. The strategy generation module obtains electricity price period information from the building energy management system to avoid executing high-energy-consuming strategies during high-electricity-price periods. The strategy adjustment module interfaces with the weather forecast system to obtain temperature and humidity forecast data 24 hours in advance and pre-adjust operating strategies to cope with weather changes. All strategy execution processes are linked to a 3D visualization model. Operators can observe the real-time impact of strategy execution on equipment status through a virtual reality interface, and click on any parameter value to view its historical change curve and related factor analysis.

[0082] Example 5: After receiving the optimized control strategy from the strategy adjustment module, the instruction generation submodule of the instruction execution module first performs an instruction compliance check to verify whether all control parameters are within the allowable operating range of the equipment. Taking the optimized strategy of an office building air conditioning system as an example, when the instruction "adjust the opening of the VAV-203 damper to 65% and set the supply air temperature to 13.5℃" is received, the system first verifies whether the damper opening is within the 0-100% effective range and whether the supply air temperature is higher than the minimum safe operating temperature of the chiller unit of 12℃. The instruction encoding unit converts the text instruction into a control protocol that the equipment can recognize, encapsulates the damper control instruction using the BACnet protocol, with the object identifier being "AV-203" and the attribute "Present-Value" set to 65; the supply air temperature instruction uses the Modbus RTU protocol, and the value written to address 40108 is 13.5 (unit: ℃). The instruction sending submodule distributes the encoded instructions to the field controllers via industrial Ethernet. The air valve control instructions are sent to the building automation system gateway, and the supply air temperature instructions are sent to the air conditioning unit PLC controller. The instruction transmission adopts a retry mechanism. If no device confirmation signal is received within 300 milliseconds, the instruction will be automatically resent up to three times.

[0083] The feedback collection submodule of the feedback update module establishes a multi-channel data acquisition link. Upon issuing a command, the feedback monitoring process is immediately initiated: it subscribes to changes in the "Actual-Value" attribute of the VAV-203 damper via the BACnet protocol and polls the real-time supply air temperature of the air conditioning unit via Modbus; simultaneously, it collects temperature and humidity data of the affected area from the environmental sensor network (such as the temperature change curve of room R-203). The feedback verification unit compares the expected effect of the command with the actual feedback. When the damper opening command is issued, the actual feedback opening value reaches 64.8% within 5 seconds (allowing ±1% error), and the supply air temperature stabilizes at 13.6℃ within 120 seconds; the temperature of room R-203 drops from 25.3℃ to 23.8℃ within 15 minutes, conforming to the expected cooling curve. The model update submodule corrects the digital twin model based on the feedback data, adjusting the actual flow characteristic parameters of the VAV-203 damper from theoretical values ​​to measured values ​​(the flow coefficient Kv is corrected from the original model value of 2.8 to 2.7), and simultaneously updates the temperature response model of the air conditioning unit under partial load. The experience base update submodule records the strategy execution effect, marks the "Summer Operating Condition Air Supply Temperature Optimization" strategy as a successful case, updates the strategy confidence score (from 0.85 to 0.88), and associates the environmental conditions (outdoor temperature 32℃ / humidity 70%) and equipment status (unit operating time 1260 hours) in the multi-layer experience base.

[0084] The command execution safety monitoring employs a dual-protection mechanism. While sending the valve opening adjustment command, the system monitors the pipeline static pressure changes in real time. When an abnormal increase in static pressure exceeding the warning value is detected, a protection command is automatically inserted: "reduce the fan frequency by 5%." Detailed operation logs are recorded for all command execution processes, including command sending time, equipment response time, actual parameter changes, and any abnormal events that occur during execution. The time synchronization accuracy of feedback data reaches the millisecond level. Each feedback data point is marked with both the device's local timestamp and the system's receiving timestamp, facilitating the analysis of command transmission delays and execution lag. The model update process uses incremental learning. The digital twin model is not a complete reconstruction but rather a parameter calibration based on the original model: the air conditioning unit's supply air temperature response model adjusts the time constant of the first-order lag element based on the current execution data (from 180 seconds to 170 seconds); the valve's flow characteristic curve updates the quadratic term coefficients based on the relationship between the actual opening and the measured airflow. The experience base update not only records successful cases but also analyzes the reasons for execution deviations. When a supply air temperature adjustment fails to achieve the expected effect, the system records possible influencing factors (such as sudden changes in fresh air temperature or increased dust accumulation in the filter), forming a knowledge graph of equipment characteristics. The closed-loop control of command execution and feedback adopts an adaptive rhythm. For fast-response equipment (such as damper actuators), a 2-second feedback timeout threshold is set; for slow-process equipment (such as room temperature control), a 15-minute effect evaluation cycle is set. All control commands carry a unique serial number, and feedback data is precisely associated with the command through the serial number, avoiding feedback confusion when multiple commands are executed concurrently. The historical command archive saves all execution records, allowing for retrospective querying of control operations on specific equipment within any time period, while also recording the operator ID and the system's automatic decision-making markers.

[0085] During system maintenance windows, automatic instruction set testing is performed, and standard test instruction sequences are sent monthly. Feedback data is analyzed to diagnose the degree of performance degradation in actuators. Test results are used to update equipment health status scores. When a damper actuator's response speed decreases by more than 15%, a maintenance work order is automatically generated. A cross-system coordination mechanism ensures the global security of instruction execution. When the energy management system indicates that the building is experiencing peak electricity consumption, non-urgent energy efficiency optimization instructions are automatically suspended. When the fire system alarms, all air conditioning control instructions are immediately overridden, and a predefined fire smoke extraction mode is executed.

[0086] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0087] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A building facility full life cycle operation and maintenance monitoring system based on digital twins, characterized in that, The system includes: The data acquisition module acquires real-time monitoring data and historical operation and maintenance data of building facilities, preprocesses and extracts features from the data, and generates standardized data. The digital twin modeling module constructs a digital twin model of the building facility based on the standardized data and simulates its operating status. The experience base formation module classifies historical operation and maintenance strategies and fault handling records based on strategy confidence and effectiveness indicators to form a multi-layered experience base. The status analysis module performs status assessment and anomaly detection based on the digital twin model and real-time monitoring data, and outputs status analysis results. The strategy generation module generates an operation and maintenance strategy plan based on the status analysis results and the multi-layer experience base. The strategy adjustment module dynamically adjusts the operation and maintenance strategy based on real-time environmental changes and equipment status, and outputs optimized control strategies. The instruction execution module generates control instructions based on the optimized control strategy and sends them to the execution device.

2. The building facility full life cycle operation and maintenance monitoring system based on digital twins according to claim 1, characterized in that, The standardized data includes data feature vectors, time-series data sets, and data quality indicators. The digital twin model includes a three-dimensional geometric model, a physical attribute model, and a behavioral simulation model. The multi-layered experience base includes strategy confidence distribution, performance index ranking, and a historical strategy set. The status analysis results include status scores, anomaly warning indicators, and predicted trend data. The operation and maintenance strategy plan includes maintenance plans, parameter adjustment, and execution steps. The optimization control strategy includes real-time instructions, parameter adjustment, and execution priority. The control instructions include equipment control commands and parameter settings. The feedback data includes execution results and equipment status changes.

3. The building facility full life cycle operation and maintenance monitoring system based on digital twins according to claim 1, characterized in that, The data acquisition module includes: The data collection submodule acquires real-time sensor data and historical operation and maintenance data of building facilities, and records the data source and timestamp; The data preprocessing submodule performs data cleaning and format conversion based on the data output by the data collection submodule, generating cleaned data; The feature extraction submodule extracts data feature vectors and time series features from the cleaned data output by the data preprocessing submodule to generate standardized data.

4. The building facility full life cycle operation and maintenance monitoring system based on digital twins according to claim 1, characterized in that, The digital twin modeling module includes: The model building submodule constructs a three-dimensional geometric model and a physical property model of the building facilities based on the standardized data; The behavior simulation submodule constructs a model based on the model output by the submodule, simulates the operational behavior of building facilities, and generates a digital twin model.

5. The building facility full life cycle operation and maintenance monitoring system based on digital twins according to claim 1, characterized in that, The experience base formation module includes: The strategy collection submodule retrieves historical operation and maintenance strategies and fault handling records; The effect evaluation submodule collects the data output by the submodule based on the strategy, and evaluates the strategy execution effect and confidence level; The classification and storage submodule classifies and stores the strategies according to the evaluation results output by the effect evaluation submodule, forming a multi-layered experience base.

6. The building facility full life cycle operation and maintenance monitoring system based on digital twins according to claim 1, characterized in that, The status analysis module includes: The data input submodule acquires the digital twin model and real-time monitoring data; The status calculation submodule calculates the status score and health level based on the data output by the data input submodule. The anomaly detection submodule performs anomaly detection and generates early warnings based on the results output by the state calculation submodule. The output module integrates status and exception information, and outputs status analysis results.

7. The building facility full life cycle operation and maintenance monitoring system based on digital twins according to claim 1, characterized in that, The strategy generation module includes: The strategy query submodule queries similar strategies from a multi-layered experience base based on the state analysis results; The strategy optimization submodule optimizes the strategy based on the queried strategy and the current status, and generates an operation and maintenance strategy plan.

8. The building facility full life cycle operation and maintenance monitoring system based on digital twins according to claim 1, characterized in that, The strategy adjustment module includes: The real-time data acquisition submodule obtains real-time environmental changes and equipment status; The dynamic adjustment submodule adjusts the operation and maintenance strategy based on the data output by the real-time data acquisition submodule, and outputs optimized control strategies.

9. The building facility full life cycle operation and maintenance monitoring system based on digital twin as described in claim 1, characterized in that, The instruction execution module includes: The instruction generation submodule generates equipment control instructions based on the optimized control strategy; The instruction sending submodule sends control instructions to the execution device.

10. The building facility full life cycle operation and maintenance monitoring system based on digital twins according to claim 1, characterized in that, The system further includes: a feedback update module, used to collect feedback data after execution and update the digital twin model and the multi-layer experience base according to the feedback data; The feedback update module includes: The feedback collection submodule collects the device status and data changes after execution; The model update submodule collects the data output by the submodule based on feedback and updates the digital twin model; The experience base update submodule updates the multi-level experience base based on feedback data and execution results.

Images