A building operation and maintenance method and system based on digital twin

Abstract

The present invention discloses a building operation and maintenance method and system based on digital twins, which relates to the field of building operation and maintenance technology, including: collecting key building equipment data, building a three-dimensional digital twin model; extracting feature data of the collected data, and using a target detection model to identify anomaly detection frames, and combining to generate a device state vector; calculating the transition probability of the state component according to the device state vector, and determining the failure probability of the equipment, and predicting the remaining life through an LSTM model; calculating the criticality score and the usage intensity score. The method of the present invention provides quantitative indicators for the frequency domain analysis of the operating status of key equipment by extracting the power distribution of the vibration signal, which is convenient for subsequent status evaluation, reflects the dynamic changes of the regional environment by calculating the rate of change of the temperature and humidity values, provides data support for analyzing the impact of the external environment on the operating status of the equipment, and contains multi-dimensional information of the equipment operating status and the external environment through the fusion vector.

Description

Technical Field

[0001] The present invention relates to the technical field of building operation and maintenance, and in particular to a building operation and maintenance method and system based on digital twins. Background Art

[0002] With the rapid advancement of modern urban construction, building operation and maintenance management has gradually become an important issue in the construction industry. Traditional building operation and maintenance methods usually rely on manual inspections and regular maintenance, which not only consumes a lot of manpower and material resources, but also has problems of low efficiency and lack of timeliness. In recent years, with the rapid development of the Internet of Things (IoT), big data analysis and artificial intelligence technology, digital operation and maintenance has become a research hotspot. Digital twin technology, as an innovative means of dynamically mapping physical equipment in virtual space, provides accurate status monitoring and prediction capabilities for building operation and maintenance by collecting data on key building equipment, combining 3D modeling and real-time data analysis.

[0003] However, the existing building operation and maintenance systems still have many shortcomings in practical applications. First, most existing technologies only realize static monitoring of equipment, lack the ability to fully perceive the dynamic status of equipment, and cannot accurately predict the operation trend of equipment. Secondly, when processing massive building equipment data, existing technologies are mostly based on single-dimensional data analysis, which makes it difficult to comprehensively consider multiple influencing factors, such as equipment operation intensity, environmental changes, and the interaction between equipment. As a result, the allocation of operation and maintenance tasks is mainly based on manual experience and lacks scientific priority sorting strategies, resulting in the processing of high-risk equipment may lag behind low-risk equipment. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a building operation and maintenance method based on digital twins to solve the problem that most existing technologies only realize static monitoring of equipment, lack the ability to comprehensively perceive the dynamic state of equipment, and cannot accurately predict the operation trend of equipment. Secondly, when processing massive building equipment data, existing technologies are mostly based on single-dimensional data analysis, which makes it difficult to comprehensively consider multiple influencing factors, such as equipment operation intensity, environmental changes, and the interaction between equipment. As a result, the allocation of operation and maintenance tasks is mainly based on manual experience, lacks scientific priority sorting strategies, and causes the processing of high-risk equipment to lag behind low-risk equipment.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a building operation and maintenance method based on digital twins, comprising:

[0008] Collect data on key building equipment and build a three-dimensional digital twin model;

[0009] Extract feature data from collected data, use target detection model to identify abnormal detection frames, and combine them to generate device state vectors;

[0010] Calculate the transition probability of the state component based on the device state vector, determine the failure probability of the device, and predict the remaining life through the LSTM model;

[0011] Calculate the criticality score and usage intensity score, analyze the equipment priority based on the remaining life of the equipment and the probability of failure, and generate an operation and maintenance list for the faulty equipment;

[0012] Count the average failure rate of building equipment, issue risk warnings, and generate warning reports.

[0013] As a preferred solution of the digital twin-based building operation and maintenance method of the present invention, wherein: the data of key building equipment is collected to construct a three-dimensional digital twin model, and three-dimensional geometric point cloud data of the exterior of the building is collected using laser scanning equipment and drones;

[0014] Scan the 3D geometric point cloud data of the internal structure of the building through laser scanning equipment;

[0015] Grid the point cloud data to generate a high-precision 3D model, which is then imported into BIM modeling software to create a building information model using parametric modeling tools;

[0016] Deploy sensors including temperature and humidity sensors, current sensors, and vibration sensors in buildings, and key equipment collects operating data in real time through edge devices;

[0017] The data collected by the sensors is transmitted to the data gateway via the IoT network through the IoT protocol MQTT. In the building information model, the operation data is synchronized in real time using data link technology;

[0018] Using the digital twin platform, the building’s external terrain data GIS is combined with the building information model to generate a three-dimensional digital twin model that includes geographic information.

[0019] As a preferred solution of the digital twin-based building operation and maintenance method of the present invention, wherein: the feature data of the collected data is extracted, and the target detection model is used to identify the abnormal detection frame, and the device state vector pointer is combined to generate the collected operation data of the key equipment, and the sliding window mean method is used to perform data denoising;

[0020] The power distribution of vibration signals collected from key equipment is extracted and expressed as:

[0021] ;

[0022] in represents the power distribution of the vibration signal at frequency f, N represents the number of sampling points, represents the signal value at sampling point k, j represents the imaginary unit;

[0023] Calculate the rate of change of the temperature and humidity values ​​of the sensors in the area and the instantaneous power of the equipment operation;

[0024] Use the target detection model YOLOV5-lite, use the target detection dataset of abnormal features of cracks and wear as the training set, and use the optimized EIoU loss function to calculate the loss value;

[0025] Based on the EIoU loss value as the objective function, the parameters of the YOLOV5-lite model are optimized through back propagation for pre-training;

[0026] Use a high-resolution camera to capture target surfaces including building walls and key equipment to obtain image data;

[0027] Normalize the image data and input it into the target detection model to perform surface feature recognition and output a set of anomaly detection boxes;

[0028] According to the set of abnormal detection frames, the areas of all abnormal detection frames are cumulatively calculated, the total area of ​​the monitoring area corresponding to the input image is calculated, and the ratio of the area of ​​all abnormal detection frames to the total area of ​​the monitoring area is used as the abnormal impact factor ;

[0029] The power distribution of the vibration signal at frequency f, the rate of change of temperature and humidity values, instantaneous power, and abnormal influencing factors are used to construct the equipment state vector, which is expressed as:

[0030] ;

[0031] in represents the device state vector at time t, Indicates the rate of change of temperature and humidity values. represents the instantaneous power at time t.

[0032] As a preferred solution of the building operation and maintenance method based on digital twins described in the present invention, the transition probability of the state component is calculated according to the device state vector, and the failure probability of the device is determined, and the remaining life is predicted by the LSTM model. The Markov chain is used to describe the dynamic changes of the building operation and maintenance state according to the device state vector, and the value of the device state vector is standardized, and the dynamic transition probability of the state is calculated, which is expressed as:

[0033] ;

[0034] in represents the transition probability of the device from the device state vector at time t to the device state vector at time t+1, n represents the characteristic number of the device state vector, Represents the device state vector medium conditional component If present, the target state component Probability of transfer;

[0035] The transition probability of the state component is further calculated based on the transition probability, which is expressed as:

[0036] ;

[0037] in Indicates that in the state vector at time t, the conditional component Under the premise of the occurrence, at time t+1 the target component The number of samples that appear, Indicates that in the state vector at time t, the conditional component The total number of occurrences, Indicates that the component in the state vector is in the conditional component The probability of transitioning from time t to time t+1, given that Indicates that time t satisfies the device state vector The jth component of Represents the device state vector at time t+1 The i-th component of ;

[0038] The failure probability of the equipment is calculated based on the calculated transition probability, which is expressed as:

[0039] ;

[0040] in represents the probability of equipment failure at time t;

[0041] based on Further expansion of the device state vector The fusion vector is obtained, which is expressed as:

[0042] ;

[0043] in represents the fusion vector at time t;

[0044] Build an LSTM model based on the LSTM neural network, including input layer, hidden layer and output layer;

[0045] Use the fused vector data of the equipment status and faults calibrated based on historical data and the remaining life labels as the training set for model training. Select the cross entropy loss function to calculate the difference between the category probability predicted by the LSTM model and the actual label. Use the back propagation algorithm to iteratively optimize and update the parameters of the LSTM model. If the model loss no longer decreases significantly during the continuous iteration process, stop iterating and output the model parameters.

[0046] Based on the trained LSTM model, the collected fusion vector is input into the model to output the remaining life data of the equipment.

[0047] As a preferred solution of the digital twin-based building operation and maintenance method described in the present invention, wherein: the criticality score and the usage intensity score are calculated, the equipment priority is analyzed in combination with the remaining life of the equipment and the probability of failure, and the equipment downtime impact duration, the number of related services and the substitutability score of the equipment are considered according to the key equipment of the building for normalization processing, and the criticality score of the key equipment is calculated;

[0048] According to the key equipment of the building, the equipment operation time, operation load ratio and operation frequency data are normalized, and the use intensity score of the key equipment is calculated;

[0049] The device priority is calculated based on the criticality score and the usage intensity score, expressed as:

[0050] ;

[0051] in represents the device priority of the i-th device, represents the failure probability of the ith device, represents the predicted remaining life of the ith device, represents the criticality score of the ith device, represents the usage intensity score of the i-th device.

[0052] As a preferred solution of the digital twin-based building operation and maintenance method of the present invention, wherein: the generation of an operation and maintenance list for judging faulty equipment refers to taking the sum of the mean and double standard deviation of the probability values ​​of historical faulty equipment as the fault threshold, if If it is greater than or equal to the fault threshold, it is considered as a device failure;

[0053] Sort the faulty devices by device priority from high to low and generate an operation and maintenance list.

[0054] As a preferred solution of the digital twin-based building operation and maintenance method described in the present invention, wherein: the failure average value of building equipment is counted, risk warning is performed, and a warning report is generated, and the failure average value of building equipment is calculated based on the failure probability value of building equipment;

[0055] The sum of the mean and standard deviation of the historical fault averages of building equipment is used as the building threshold. If the fault average is greater than or equal to the building threshold, the building is judged to have a high risk of failure and a building maintenance warning is triggered.

[0056] Send emergency warning information to the maintenance team, including building number, average failure value, and remaining life data, and generate warning reports through report generation tools.

[0057] In a second aspect, the present invention provides a system for a building operation and maintenance method based on digital twins, comprising:

[0058] The 3D digital twin construction module collects real-time data of key building equipment and uses laser scanning, drones, and BIM modeling technology to build a 3D digital twin model;

[0059] The device state vector generation module extracts the features of the collected data, uses the target detection model to identify anomalies, and generates the device state vector;

[0060] The equipment status dynamic analysis module calculates the dynamic transition probability of the state component based on the equipment status vector, determines the equipment failure probability, and predicts the remaining life;

[0061] The equipment priority analysis module calculates the criticality score and usage intensity score, combines the remaining life and failure probability, analyzes the equipment priority, and generates an operation and maintenance list;

[0062] The risk warning module counts the average failure rate of building equipment, issues risk warnings, and generates warning reports based on building failure risks.

[0063] The beneficial effects of the present invention are as follows: by extracting the power distribution of vibration signals, quantitative indicators are provided for the frequency domain analysis of the operating status of key equipment, which is convenient for subsequent status evaluation; by calculating the rate of change of temperature and humidity values, the dynamic changes of the regional environment are reflected, and data support is provided for analyzing the impact of the external environment on the operating status of the equipment; by fusion vectors containing multi-dimensional information of the equipment operating status and the external environment, a comprehensive and in-depth description of the equipment status is achieved; by introducing the probability of failure, a systematic risk assessment is provided; by combining environmental and operating data, the accuracy and completeness of the equipment status representation are significantly improved; by combining the real-time input of the fusion vector with the LSTM model during the operation of the equipment, the remaining life prediction can be dynamically updated, providing continuous intelligent support for equipment operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0065] Figure 1 This is a flow chart of the building operation and maintenance method based on digital twins in Example 1.

[0066] Figure 2 This is a structural diagram of the building operation and maintenance system based on digital twins in Example 1. DETAILED DESCRIPTION

[0067] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0068] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0069] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0070] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a building operation and maintenance method based on digital twins, comprising the following steps:

[0071] S1, collects data of key building equipment and builds a three-dimensional digital twin model;

[0072] Preferably, data of key building equipment is collected, a three-dimensional digital twin model is constructed, and three-dimensional geometric point cloud data of the exterior of the building is collected using laser scanning equipment and drones;

[0073] Scan the 3D geometric point cloud data of the internal structure of the building through laser scanning equipment;

[0074] Grid the point cloud data to generate a high-precision 3D model, which is then imported into BIM modeling software to create a building information model using parametric modeling tools;

[0075] Deploy sensors including temperature and humidity sensors, current sensors, and vibration sensors in buildings and key equipment (including air conditioners, boilers, and elevators) to collect real-time operating data through edge devices;

[0076] The data collected by the sensors is transmitted to the data gateway via the IoT network through the IoT protocol MQTT. In the building information model, the operation data is synchronized in real time using data link technology;

[0077] Use a digital twin platform (such as Unity Reflect) to combine the building's external terrain data GIS with the building information model to generate a three-dimensional digital twin model that includes geographic information.

[0078] Through the gridding of high-precision point cloud data, the constructed 3D model can accurately reflect the actual structure of the building, eliminating the deviation between traditional design and construction. Laser scanning and drone technology support the rapid update of model data, so that the building information can be adjusted in real time as the physical state of the building changes. The combination of BIM software's parametric modeling tools and gridding data improves the compatibility of the model and facilitates docking with other building systems.

[0079] Real-time collection of the operating status of key equipment supports accurate fault location and prediction. Through the MQTT protocol, data transmission has low latency and high reliability, supporting edge computing devices to quickly process large amounts of sensor data. By synchronizing operating data in real time, the three-dimensional digital twin model can dynamically display the operating status of building equipment. The digital twin platform integrates BIM and GIS data, which can not only display the internal structure of the building, but also analyze the overall environmental impact in combination with geographic information.

[0080] S2, extracts the feature data of the collected data, and uses the target detection model to identify the abnormal detection box, and combines them to generate the device state vector;

[0081] Preferably, feature data of the collected data is extracted, and an abnormal detection frame is identified using a target detection model, and the device state vector pointer is combined to generate the collected operation data of the key equipment, and the sliding window mean method is used to perform data denoising;

[0082] The power distribution of vibration signals collected from key equipment is extracted and expressed as:

[0083] ;

[0084] in represents the power distribution of the vibration signal at frequency f, N represents the number of sampling points, represents the signal value at sampling point k, j represents the imaginary unit;

[0085] Calculate the rate of change of the temperature and humidity values ​​of the sensors in the area and the instantaneous power of the equipment operation;

[0086] The target detection model YOLOV5-lite is used, the target detection dataset of abnormal features of cracks and wear is used as the training set, and the loss value is calculated using the optimized EIoU loss function, which is expressed as:

[0087] ;

[0088] in Represents the EIoU loss value, It represents the overlap intersection ratio between the detection box and the real box. Represents the Euclidean distance between the center point of the detection box A and the real box B, Indicates the diagonal length of the minimum bounding box;

[0089] Based on the EIoU loss value as the objective function, the parameters of the YOLOV5-lite model are optimized through back propagation for pre-training;

[0090] Use a high-resolution camera to capture target surfaces including building walls and key equipment to obtain image data;

[0091] Normalize the image data and input it into the target detection model to perform surface feature recognition and output a set of anomaly detection boxes;

[0092] According to the set of abnormal detection frames, the areas of all abnormal detection frames are cumulatively calculated, the total area of ​​the monitoring area corresponding to the input image is calculated, and the ratio of the area of ​​all abnormal detection frames to the total area of ​​the monitoring area is used as the abnormal impact factor ;

[0093] The power distribution of the vibration signal at frequency f, the rate of change of temperature and humidity values, instantaneous power, and abnormal influencing factors are used to construct the equipment state vector, which is expressed as:

[0094] ;

[0095] in represents the device state vector at time t, Indicates the rate of change of temperature and humidity values. represents the instantaneous power at time t.

[0096] Denoising by the sliding window mean method can smooth random fluctuations in the data, reduce noise interference in the collected data, and improve data accuracy. By extracting the power distribution of the vibration signal, quantitative indicators are provided for the frequency domain analysis of the operating status of key equipment, which is convenient for subsequent status evaluation. The dynamic changes of the regional environment are reflected by calculating the change rate of the temperature and humidity values, providing data support for analyzing the impact of the external environment on the operating status of the equipment. The actual workload of the equipment is reflected by the instantaneous power, supporting real-time detection of abnormal operation of the equipment. The optimized EIoU loss function takes into account the box position, shape and center point distance, which improves the accuracy of the target detection model in surface anomaly detection. The combination of high-resolution cameras and target detection models can quickly locate surface anomalies of building walls and key equipment, improving detection efficiency. The abnormal influencing factors provide a data basis for the comprehensive evaluation of regional anomalies, supporting the priority sorting of operation and maintenance tasks, and the state vector integrates the internal operating characteristics of the equipment and the impact of the external environment to achieve multi-dimensional characterization of the equipment status.

[0097] S3, calculates the transition probability of the state component according to the device state vector, determines the failure probability of the device, and predicts the remaining life through the LSTM model;

[0098] Preferably, the transition probability of the state component is calculated according to the equipment state vector, and the failure probability of the equipment is determined. The remaining life is predicted by the LSTM model. The Markov chain is used to describe the dynamic changes of the building operation and maintenance state according to the equipment state vector. The value of the equipment state vector is standardized and the dynamic transition probability of the state is calculated, which is expressed as:

[0099] ;

[0100] in represents the transition probability of the device from the device state vector at time t to the device state vector at time t+1, n represents the characteristic number of the device state vector (which can be 4 here), Represents the device state vector medium conditional component If present, the target state component The probability of transfer is based on the conditional probability obtained from historical operation data statistics. Represents the device state vector The state transition probability of each feature component in The product of is used to describe the overall state transition of the system;

[0101] The transition probability of the state component is further calculated based on the transition probability, which is expressed as:

[0102] ;

[0103] in Indicates that in the state vector at time t, the conditional component Under the premise of the occurrence, at time t+1 the target component The number of samples that appear, Indicates that in the state vector at time t, the conditional component The total number of occurrences, Indicates that the component in the state vector is in the conditional component The probability of transitioning from time t to time t+1, given that Indicates that time t satisfies the device state vector The jth component of Represents the device state vector at time t+1 The i-th component of ;

[0104] The failure probability of the equipment is calculated based on the calculated transition probability, which is expressed as:

[0105] ;

[0106] in represents the probability of equipment failure at time t;

[0107] based on Further expansion of the device state vector The fusion vector is obtained, which is expressed as:

[0108] ;

[0109] in represents the fusion vector at time t;

[0110] Build an LSTM model based on the LSTM neural network, including input layer, hidden layer and output layer;

[0111] Use the fused vector data of the equipment status and faults calibrated based on historical data and the remaining life labels as the training set for model training. Select the cross entropy loss function to calculate the difference between the category probability predicted by the LSTM model and the actual label. Use the back propagation algorithm to iteratively optimize and update the parameters of the LSTM model. If the model loss no longer decreases significantly during the continuous iteration process, stop iterating and output the model parameters.

[0112] Based on the trained LSTM model, the collected fusion vector is input into the model to output the remaining life data of the equipment.

[0113] Through the Markov chain, the changing patterns of the equipment operating status over time are dynamically captured, providing a basis for subsequent predictions. The transition probability is used to quantitatively describe the changes in the equipment status, which facilitates the discovery of potential correlations between key states. The fault probability model provides quantitative results of the fault risk of the current operating status of the equipment. According to the fault probability, the abnormal status that may cause equipment shutdown is quickly identified, providing a basis for operation and maintenance decisions. The fusion vector integrates the equipment status characteristics and fault risks, providing comprehensive data support for the comprehensive evaluation of the equipment status. The LSTM model can capture the long-sequence dependency characteristics of the equipment operating status, laying the foundation for accurate prediction of the remaining life.

[0114] By integrating the multi-dimensional information of the equipment operation status and external environment into the fusion vector, a comprehensive and in-depth description of the equipment status is achieved. The introduction of the probability of failure provides a systematic risk assessment. By combining the environment and operation data, the accuracy and completeness of the equipment status representation are significantly improved. Through the fusion vector, the subsequent LSTM model can be trained based on richer data features to optimize the effect of the remaining life prediction. During the operation of the equipment, the real-time input of the fusion vector is combined with the LSTM model, so that the remaining life prediction can be dynamically updated, providing continuous intelligent support for equipment operation and maintenance. Based on accurate remaining life prediction, equipment managers can scientifically formulate maintenance plans to avoid waste of resources or sudden failures caused by premature or late maintenance.

[0115] Through accurate prediction capabilities and priority sorting, unnecessary maintenance and inspections can be reduced, and the manpower and material costs of operation and maintenance can be reduced. The comprehensive evaluation of failure probability and remaining life can avoid waste of resources and form a scientific operation and maintenance plan. The combination of dynamic state transition probability, failure probability and fusion vector, and intelligent prediction through LSTM model have broken through the limitations of traditional single-dimensional operation and maintenance methods. The introduction of new features (such as historical equipment load, external climate conditions, etc.) is supported through system architecture design to further improve prediction and decision-making capabilities. The Markov chain, LSTM neural network and Industrial Internet of Things (IIoT) data are combined to build intelligent operation and maintenance solutions for complex systems.

[0116] S4, calculates the criticality score and usage intensity score, analyzes the equipment priority based on the remaining life of the equipment and the probability of failure, determines the faulty equipment and generates an operation and maintenance list;

[0117] Preferably, the criticality score and the usage intensity score are calculated, and the equipment priority is analyzed in combination with the remaining life of the equipment and the probability of failure. The impact time of equipment downtime, the number of related services, and the substitutability score of the equipment are considered according to the key equipment of the building for normalization, and the criticality score of the key equipment is calculated, which is expressed as:

[0118] ;

[0119] in represents the criticality score, Indicates the duration of equipment downtime. Indicates the maximum downtime impact duration in the device. Indicates the number of services associated with the device. Indicates the maximum number of associated services in the device. represents the substitutability score of the device, Indicates the maximum value of the substitutability score of the device;

[0120] According to the key equipment of the building, the equipment operation time, operation load proportion and operation frequency data are normalized, and the use intensity score of the key equipment is calculated, which is expressed as:

[0121] ;

[0122] in Indicates the usage intensity score, Indicates the current cumulative running time of the device. Indicates the maximum cumulative running time of all devices. Indicates the proportion of the equipment's operating load. Indicates the maximum operating load ratio among all devices. Indicates the number of times the device starts running within a specified time. Indicates the maximum operating frequency among all devices;

[0123] The device priority is calculated based on the criticality score and the usage intensity score, expressed as:

[0124] ;

[0125] in represents the device priority of the i-th device, represents the failure probability of the ith device, represents the predicted remaining life of the ith device, represents the criticality score of the ith device, represents the usage intensity score of the i-th device.

[0126] The criticality score is used to quantify the impact of equipment on building operations and business, providing a basis for identifying equipment with high business dependence. The intensity score is used to quantify the current operating load and stress level of the equipment, which can identify high-load, high-frequency use equipment. The priority calculation combines the equipment's failure probability, remaining life, business importance and operating intensity to provide a multi-dimensional comprehensive assessment of equipment status, giving priority to equipment with a high failure probability, short remaining life and large business impact, ensuring efficient use of resources and priority resolution of key issues. The priority is dynamically adjusted based on real-time updated status data to adapt to changes in equipment status and ensure the accuracy of predictions and decisions.

[0127] Furthermore, judging the faulty equipment to generate the operation and maintenance list refers to taking the sum of the probability value mean and double standard deviation of the historical faulty equipment as the fault threshold. If it is greater than or equal to the fault threshold, it is considered as a device failure;

[0128] Sort the faulty devices by device priority from high to low and generate an operation and maintenance list.

[0129] The fault threshold based on statistical analysis dynamically adapts to the characteristics of equipment failures, avoiding overly strict or loose threshold judgments. The combination of mean and twice the standard deviation can effectively distinguish between normal status and fault status. The dynamically adjusted fault threshold improves the accuracy of judgment and reduces the probability of false alarms (normal equipment is misjudged as faulty) and missed alarms (faulty equipment is not identified). The operation and maintenance list can give priority to high-risk and high-business-impact equipment to ensure that resources are used to solve key problems. The operation and maintenance list can be dynamically adjusted according to the real-time updated priority to adapt to changes in equipment status and maintain the flexibility of the operation and maintenance plan. By combining dynamic threshold judgment with real-time priority sorting, key equipment with potential risks in building operation can be quickly identified. The operation and maintenance list sorting clarifies the priority of equipment maintenance, so that human, material and time resources can be efficiently allocated, high-risk equipment can be handled in advance, the probability of unplanned downtime caused by sudden failures can be reduced, and the reliability of building equipment operation can be improved.

[0130] S5, count the average failure rate of building equipment, conduct risk warning, and generate warning reports;

[0131] Preferably, the average failure rate of building equipment is counted, risk warning is performed, and a warning report is generated. Based on the failure probability value of the building equipment, the average failure rate of the building equipment is calculated;

[0132] The sum of the mean and standard deviation of the historical fault averages of building equipment is used as the building threshold. If the fault average is greater than or equal to the building threshold, the building is judged to have a high risk of failure and a building maintenance warning is triggered.

[0133] Send emergency warning information to the maintenance team, including building number, average failure value, and remaining life data, and generate warning reports through report generation tools.

[0134] Through the threshold judgment of the fault average value and historical data, a quantitative building fault risk assessment method is provided to avoid the limitations of single device fault judgment. The fault risk assessment based on statistical methods makes the early warning trigger mechanism more scientific and reduces false alarms and missed alarms. The risk early warning trigger mechanism can quickly identify high-risk buildings and send emergency warning information to the maintenance team in real time. The report generation tool converts complex data into intuitive warning reports to improve the readability and dissemination efficiency of information.

[0135] This embodiment also provides a system for a building operation and maintenance method based on digital twins, comprising:

[0136] The 3D digital twin construction module collects real-time data of key building equipment and uses laser scanning, drones, and BIM modeling technology to build a 3D digital twin model;

[0137] The device state vector generation module extracts the features of the collected data, uses the target detection model to identify anomalies, and generates the device state vector;

[0138] The equipment status dynamic analysis module calculates the dynamic transition probability of the state component based on the equipment status vector, determines the equipment failure probability, and predicts the remaining life;

[0139] The equipment priority analysis module calculates the criticality score and usage intensity score, combines the remaining life and failure probability, analyzes the equipment priority, and generates an operation and maintenance list;

[0140] The risk warning module counts the average failure rate of building equipment, issues risk warnings, and generates warning reports based on building failure risks.

[0141] In summary, the present invention provides quantitative indicators for frequency domain analysis of the operating status of key equipment by extracting the power distribution of vibration signals, which is convenient for subsequent status evaluation. It reflects the dynamic changes of the regional environment by calculating the rate of change of temperature and humidity values, and provides data support for analyzing the impact of the external environment on the operating status of the equipment. The fusion vector contains multi-dimensional information of the equipment operating status and the external environment, so as to achieve a comprehensive and in-depth description of the equipment status. The introduction of failure probability provides a systematic risk assessment, and the accuracy and completeness of the equipment status representation are significantly improved by combining environmental and operating data. During the operation of the equipment, the real-time input of the fusion vector is combined with the LSTM model, so that the remaining life prediction can be dynamically updated, providing continuous intelligent support for equipment operation and maintenance.

[0142] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A building operation and maintenance method based on digital twins, characterized in that: include: Deploy sensors to collect data from key building equipment, including air conditioners, boilers, and elevators. Deploy sensors include temperature and humidity sensors, current sensors, and vibration sensors to build a three-dimensional digital twin model. Extract the vibration signal power distribution of the collected data and calculate the rate of change of the temperature and humidity values ​​of the sensors in the area and the instantaneous power of the equipment operation. Use the target detection model to identify the abnormal features of cracks and wear, and identify the abnormal detection frame. The ratio of the area of ​​the abnormal detection frame to the total area of ​​the monitoring area is used as the abnormal influencing factor. The vibration signal power distribution, the rate of change of the temperature and humidity values, the instantaneous power and the abnormal influencing factor are combined to generate the equipment state vector. According to the equipment state vector, the dynamic changes of the building operation and maintenance status are identified according to the transition probability of the state vector at different times. Based on the dynamic changes identified by the transition probability, the ratio of the number of target components under different conditional components of the state vector to the total number of conditional components is determined as the dynamic transition probability. The failure probability of the equipment is further calculated based on the dynamic transition probability. Based on the calculation of the failure probability, the state vector is further expanded to obtain the fusion vector, and the remaining life is predicted through the LSTM model. Calculate the criticality score and usage intensity score, analyze the equipment priority based on the remaining life of the equipment and the probability of failure, and generate an operation and maintenance list for the faulty equipment; Count the average failure rate of building equipment, conduct risk warnings, and generate warning reports; The power distribution of vibration signals collected from key equipment is extracted and expressed as: ； in represents the power distribution of the vibration signal at frequency f, N represents the number of sampling points, represents the signal value at sampling point k, j represents the imaginary unit; Calculate the rate of change of the temperature and humidity values ​​of the sensors in the area and the instantaneous power of the equipment operation; Use the target detection model YOLOV5-lite, use the target detection dataset of abnormal features of cracks and wear as the training set, and use the optimized EIoU loss function to calculate the loss value; Based on the EIoU loss value as the objective function, the parameters of the YOLOV5-lite model are optimized through back propagation for pre-training; Use a high-resolution camera to capture target surfaces including building walls and key equipment to obtain image data; Normalize the image data and input it into the target detection model to perform surface feature recognition and output a set of anomaly detection boxes; According to the set of abnormal detection frames, the areas of all abnormal detection frames are cumulatively calculated, the total area of ​​the monitoring area corresponding to the input image is calculated, and the ratio of the area of ​​all abnormal detection frames to the total area of ​​the monitoring area is used as the abnormal impact factor ; The power distribution of the vibration signal at frequency f, the rate of change of temperature and humidity values, instantaneous power, and abnormal influencing factors are used to construct the equipment state vector, which is expressed as: ； in represents the device state vector at time t, Indicates the rate of change of temperature and humidity values. represents the instantaneous power at time t.

2. The building operation and maintenance method based on digital twins according to claim 1, characterized in that: The key equipment data of the building is collected, a three-dimensional digital twin model is constructed, and laser scanning equipment and drones are used to collect three-dimensional geometric point cloud data outside the building; Scan the 3D geometric point cloud data of the internal structure of the building through laser scanning equipment; Grid the point cloud data to generate a high-precision 3D model, which is then imported into BIM modeling software to create a building information model using parametric modeling tools; Deploy sensors including temperature and humidity sensors, current sensors, and vibration sensors in buildings, and key equipment collects operating data in real time through edge devices; The data collected by the sensors is transmitted to the data gateway via the IoT network through the IoT protocol MQTT. In the building information model, the operation data is synchronized in real time using data link technology; Using the digital twin platform, the building’s external terrain data GIS is combined with the building information model to generate a three-dimensional digital twin model that includes geographic information.

3. The building operation and maintenance method based on digital twins according to claim 2, characterized in that: The feature data of the collected data is extracted, and the target detection model is used to identify the abnormal detection frame, and the device state vector pointer is combined to generate the collected operation data of the key equipment, and the sliding window mean method is used to perform data denoising.

4. The building operation and maintenance method based on digital twins according to claim 3, characterized in that: The transition probability of the state component is calculated according to the device state vector, and the failure probability of the device is determined. The remaining life is predicted through the LSTM model. The Markov chain is used to describe the dynamic changes of the building operation and maintenance status according to the device state vector, and the value of the device state vector is standardized to calculate the dynamic transition probability of the state, which is expressed as: ； in represents the transition probability of the device from the device state vector at time t to the device state vector at time t+1, n represents the characteristic number of the device state vector, Represents the device state vector medium conditional component If present, the target state component Probability of transfer; The transition probability of the state component is further calculated based on the transition probability, which is expressed as: ； in Indicates that in the state vector at time t, the conditional component Under the premise of the appearance, at time t+1 the target component The number of samples that appear, Indicates that in the state vector at time t, the conditional component The total number of occurrences, Indicates that the component in the state vector is in the conditional component The probability of transitioning from time t to time t+1, given that Indicates that time t satisfies the device state vector The jth component of Represents the device state vector at time t+1 The i-th component of ; The failure probability of the equipment is calculated based on the calculated transition probability, which is expressed as: ； in represents the probability of equipment failure at time t; based on Further expansion of the device state vector The fusion vector is obtained, expressed as: ； in represents the fusion vector at time t; Build an LSTM model based on the LSTM neural network, including input layer, hidden layer and output layer; Use the fused vector data of the equipment status and faults calibrated based on historical data and the remaining life labels as the training set for model training. Select the cross entropy loss function to calculate the difference between the category probability predicted by the LSTM model and the actual label. Use the back propagation algorithm to iteratively optimize and update the parameters of the LSTM model. If the model loss no longer decreases significantly during the continuous iteration process, stop iterating and output the model parameters. Based on the trained LSTM model, the collected fusion vector is input into the model to output the remaining life data of the equipment.

5. The building operation and maintenance method based on digital twins according to claim 4, characterized in that: The criticality score and usage intensity score are calculated, and the equipment priority is analyzed in combination with the remaining life of the equipment and the probability of failure. The criticality score of the key equipment is calculated by considering the impact time of equipment downtime, the number of related services and the substitutability score of the equipment according to the key equipment of the building, and the criticality score of the key equipment is calculated; According to the key equipment of the building, the equipment operation time, operation load ratio and operation frequency data are normalized, and the use intensity score of the key equipment is calculated; The device priority is calculated based on the criticality score and the usage intensity score, expressed as: ； in represents the device priority of the i-th device, represents the failure probability of the ith device, represents the predicted remaining life of the ith device, represents the criticality score of the ith device, represents the usage intensity score of the i-th device.

6. The digital twin-based building operation and maintenance method according to claim 5, characterized in that: The generation of the operation and maintenance list for judging the faulty equipment refers to taking the sum of the mean and double standard deviation of the probability values ​​of the historical faulty equipment as the fault threshold. If it is greater than or equal to the fault threshold, it is considered as a device failure; Sort the faulty devices by device priority from high to low and generate an operation and maintenance list.

7. The building operation and maintenance method based on digital twins according to claim 6, characterized in that: The method of calculating the average failure rate of building equipment, conducting risk warning, and generating a warning report is to calculate the average failure rate of building equipment based on the failure probability value of the building equipment; The sum of the mean and standard deviation of the historical fault averages of building equipment is used as the building threshold. If the fault average is greater than or equal to the building threshold, the building is judged to have a high risk of failure and a building maintenance warning is triggered. Send emergency warning information to the maintenance team, including building number, average failure value, and remaining life data, and generate warning reports through report generation tools.

8. A system for a building operation and maintenance method based on digital twins, based on the building operation and maintenance method based on digital twins according to any one of claims 1 to 7, characterized in that: include, The 3D digital twin construction module collects real-time data of key building equipment and uses laser scanning, drones, and BIM modeling technology to build a 3D digital twin model; The device state vector generation module extracts the features of the collected data, uses the target detection model to identify anomalies, and generates the device state vector; The equipment status dynamic analysis module calculates the dynamic transition probability of the state component based on the equipment status vector, determines the equipment failure probability, and predicts the remaining life; The equipment priority analysis module calculates the criticality score and usage intensity score, combines the remaining life and failure probability, analyzes the equipment priority, and generates an operation and maintenance list; The risk warning module counts the average failure rate of building equipment, issues risk warnings, and generates warning reports based on building failure risks.

Images