Building equipment full life cycle management method and system

Abstract

The invention relates to a building equipment full life cycle management method and system, and belongs to the technical field of building equipment management.The management method comprises the steps that design parameters, a historical accident data set and environment reference data of building equipment are acquired; constructing a risk prediction model, generating a dynamic threshold rule base, and outputting a reference digital model; deploying a multi-modal sensor group, carrying out noise filtering and credibility weight assignment on initially acquired data, and outputting a calibration sensor network data set; calling a reference threshold value, and calculating a real-time early warning threshold value in combination with external environment dynamic data; performing weighted aggregation on the data flow according to the credibility weight, activating a risk prediction model to generate a risk level signal when the aggregated data exceeds a real-time early warning threshold, and outputting a real-time health status report and an abnormal event data set; and predicting residual life probability distribution through the equipment degradation model, and outputting a predictive maintenance work order. According to the invention, the intelligent level and the safety guarantee capability of building equipment management are improved.

Description

Technical Field

[0001] This application relates to the field of building equipment management technology, and in particular to a method and system for the full life cycle management of building equipment. Background Technology

[0002] In the context of current technology, with the acceleration of urbanization and the continuous expansion of building scale, the number of various equipment in buildings is growing rapidly and their functions are becoming increasingly complex. This places higher demands on the safe operation, efficient management, and long-term stability of equipment. Traditional building equipment management models usually rely on manual inspections and regular maintenance. This approach has obvious lag and subjectivity, making it difficult to detect potential faults in a timely manner, which can easily lead to sudden shutdowns or even safety accidents.

[0003] In recent years, some smart building projects have introduced sensor networks and remote monitoring methods, but most remain at the stage of status display and post-event alarms, failing to form an effective risk prediction capability and a forward-looking maintenance strategy support system. Especially in risk assessment, existing systems do not fully consider the impact of environmental changes on equipment performance, nor do they effectively utilize historical operating data for in-depth trend analysis. In actual operation, they also face a series of complex problems such as difficulties in fusing multi-source heterogeneous data, sensor accuracy drift, and frequent on-site interference signals, further hindering the improvement of intelligent management levels. Summary of the Invention

[0004] To improve the intelligence level and security capabilities of building equipment management, this application provides a method and system for the full life cycle management of building equipment.

[0005] Firstly, this application provides a method for the full lifecycle management of building equipment, employing the following technical solution: A method for full lifecycle management of building equipment, the method comprising: Obtain building equipment design parameters, historical accident datasets, and environmental baseline data; Based on the design parameters, historical accident datasets, and environmental benchmark data, a risk prediction model is constructed using machine learning algorithms, and a dynamic threshold rule base is generated, outputting a benchmark digital model containing the unique identifier of the equipment. Based on the aforementioned benchmark digital model, a multimodal sensor group is deployed on the physical structure of the building equipment. Each sensor is bound to the unique identifier of the equipment, and noise filtering and confidence weight assignment are performed on the initial data collected by the multimodal sensor group to output a calibration sensor network dataset. Real-time acquisition of the data stream of the calibration sensor network dataset and dynamic data of the external environment; The baseline threshold in the dynamic threshold rule base is called, and the real-time early warning threshold is calculated by combining the dynamic data of the external environment. The data stream is weighted and aggregated according to the credibility weight, and the aggregated data is compared with the real-time warning threshold. When the aggregated data exceeds the real-time warning threshold, the risk prediction model is activated to generate a risk level signal and output a real-time health status report and abnormal event dataset. Based on the real-time health status report and maintenance history, the remaining lifespan probability distribution is predicted through the equipment degradation model, and predictive maintenance work orders are output.

[0006] By adopting the above technical solutions, a complete technical chain from data acquisition to intelligent decision-making is constructed, realizing digital control over the entire process of building equipment from design, operation, maintenance to decommissioning. The core of this solution lies in establishing a risk prediction mechanism and dynamic threshold adjustment system based on machine learning. Through multimodal sensor networks, it realizes comprehensive real-time monitoring of building equipment status, and on this basis, it forms a closed-loop management system from anomaly detection to predictive maintenance, significantly improving the intelligence level and safety assurance capabilities of building equipment management.

[0007] Secondly, this application provides a building equipment full lifecycle management system, which adopts the following technical solution: A building equipment lifecycle management system, the management system comprising: The acquisition module is used to acquire the design parameters of building equipment, historical accident datasets, and environmental baseline data; The model building module is used to build a risk prediction model based on the design parameters, historical accident dataset and environmental benchmark data, through machine learning algorithms, generate a dynamic threshold rule base, and output a benchmark digital model containing the unique identifier of the equipment. The sensor configuration module is used to deploy a multimodal sensor group on the physical structure of building equipment based on the benchmark digital model, bind each sensor to the unique identifier of the equipment, perform noise filtering and confidence weight assignment on the initial acquisition data of the multimodal sensor group, and output a calibration sensor network dataset. The acquisition module is used to acquire the data stream of the calibration sensor network dataset and the dynamic data of the external environment in real time; The warning threshold calculation module is used to call the baseline threshold in the dynamic threshold rule library and combine it with the dynamic data of the external environment to calculate the real-time warning threshold. The aggregation and comparison module is used to perform weighted aggregation on the data stream according to the credibility weight, and compare the aggregated data with the real-time warning threshold. The risk prediction module is used to activate the risk prediction model to generate a risk level signal and output a real-time health status report and abnormal event dataset when the aggregated data exceeds the real-time warning threshold. The maintenance work order generation module is used to predict the probability distribution of remaining lifespan based on the real-time health status report and maintenance history, and output predictive maintenance work orders.

[0008] Thirdly, this application provides a computer device, which adopts the following technical solution: A computer device includes a memory, a processor, and a computer program stored in the memory, the processor executing the computer program to perform the steps of the method as described in the first aspect.

[0009] Fourthly, this application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium storing a computer program that can be loaded by a processor and executed as in any of the methods in the first aspect. Attached Figure Description

[0010] Figure 1 This is a first flowchart of a building equipment lifecycle management method according to one embodiment of this application.

[0011] Figure 2 This is a second flowchart of a building equipment lifecycle management method according to one embodiment of this application.

[0012] Figure 3 This is a schematic diagram of the third process of a building equipment lifecycle management method according to one embodiment of this application.

[0013] Figure 4 This is a schematic diagram of the fourth process of a building equipment lifecycle management method according to one embodiment of this application.

[0014] Figure 5 This is a schematic diagram of the fifth process of a building equipment lifecycle management method according to one embodiment of this application.

[0015] Figure 6 This is a schematic diagram of the sixth process of a building equipment lifecycle management method according to one embodiment of this application.

[0016] Figure 7 This is a schematic diagram of the seventh process of a building equipment lifecycle management method according to one embodiment of this application. Detailed Implementation

[0017] To make the purpose, technical solution, and advantages of this application clearer, the following description is provided in conjunction with the appendix. Figure 1-7The present application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the application.

[0018] This application discloses a method for the full lifecycle management of building equipment.

[0019] Reference Figure 1 A method for managing the entire lifecycle of building equipment, the method including: Step S101: Obtain the design parameters of building equipment, historical accident datasets, and environmental baseline data; The design parameters encompass core engineering and technical indicators such as the equipment's structural specifications, material properties, and load-bearing capacity. These parameters determine the equipment's basic performance boundaries and theoretical operating range. Historical accident datasets contain valuable empirical data, including past fault records, failure modes, and maintenance history. In-depth analysis of this historical data can identify vulnerable parts and common risk points within the equipment. Environmental benchmark data provides long-term statistical characteristics of the equipment's geographical environment, including external factors such as geological conditions, climate parameters, and load characteristics. This data lays the foundation for establishing an adaptive early warning mechanism. This data acquisition process essentially provides multi-dimensional training samples and references for subsequent risk modeling, ensuring that the model can comprehensively consider the interaction between the equipment's inherent characteristics and the external environment.

[0020] Step S102: Based on design parameters, historical accident datasets and environmental benchmark data, a risk prediction model is constructed using machine learning algorithms, and a dynamic threshold rule base is generated, outputting a benchmark digital model containing the equipment's unique identifier. The process of building the risk prediction model essentially involves establishing a nonlinear mapping relationship between equipment status parameters and their potential risks through pattern recognition and statistical analysis of historical accident data. This enables the system to promptly identify and quantify the risk level when abnormal signs appear in the equipment.

[0021] Furthermore, traditional fixed thresholds often fail to adapt to complex and ever-changing actual working conditions. Therefore, it is necessary to establish a threshold system that can dynamically adjust according to environmental conditions. This dynamism is reflected in the fact that the threshold is no longer a single value but a function related to environmental parameters, i.e., generating a dynamic threshold rule base. The benchmark digital model, as the data hub of the entire system, not only contains the geometric information and attribute characteristics of the equipment, but more importantly, the embedded unique equipment identifier provides a unified identity authentication mechanism for the association and tracking of all subsequent data, ensuring end-to-end traceability from sensor data to maintenance instructions.

[0022] Step S103: Based on the benchmark digital model, deploy a multimodal sensor group on the physical structure of the building equipment, bind each sensor to a unique device identifier, and perform noise filtering and confidence weight assignment on the initial collected data of the multimodal sensor group to output a calibration sensor network dataset. Based on the obtained baseline digital model, the system enters the physical deployment phase, which involves deploying multimodal sensor arrays on the physical structure of the building equipment and binding each sensor to a unique device identifier. The concept of multimodal sensor arrays reflects the development trend of modern monitoring technology. Different types of sensors can capture different physical state information of the equipment; the coordinated monitoring of multiple parameters such as strain, vibration, temperature, and humidity can more comprehensively reflect the overall health status of the equipment. The process of binding sensors to unique device identifiers effectively establishes a mapping relationship between the physical and digital worlds, ensuring that the data collected by each sensor can be accurately attributed to a specific part of a specific device. This precise binding mechanism provides a reliable foundation for subsequent data fusion and fault location.

[0023] Subsequently, noise filtering and confidence weight assignment were performed on the initial acquired data. Noise filtering removed interference components through signal processing techniques, while confidence weight assignment assigned corresponding confidence levels to different sensor data based on the signal quality assessment results. This differentiated processing ensured the accuracy of subsequent data analysis.

[0024] Step S104: Real-time acquisition of the data stream of the calibration sensor network dataset and dynamic data of the external environment; Once the system enters the real-time monitoring phase, it continuously acquires data streams from the calibration sensor network dataset and dynamic data from the external environment. Continuous acquisition of the real-time data stream ensures the system's sensitivity to changes in device status, enabling it to promptly detect even minor anomalies. Simultaneous acquisition of dynamic external environmental data provides the necessary input parameters for dynamic threshold adjustments. This architecture, with its parallel acquisition of dual data streams, reflects the system's forward-thinking design. The normal operating range of the device is not static but adjusts according to changes in environmental conditions; only by simultaneously understanding the real-time information of both the device's internal state and the external environment can accurate status judgments be made.

[0025] Step S105: Call the baseline threshold in the dynamic threshold rule base and calculate the real-time warning threshold in combination with the dynamic data of the external environment; The core of this process lies in transforming static threshold rules into dynamic early warning standards. Environmental parameters are used to adjust the baseline thresholds, making them adaptable to specific operating conditions. For example, in high-temperature environments, the properties of certain materials may change, requiring adjustments to the corresponding safety thresholds. This adaptive threshold calculation mechanism significantly improves the accuracy and practicality of the early warning system.

[0026] Step S106: Perform weighted aggregation of the data stream according to the credibility weight, and compare the aggregated data with the real-time early warning threshold; The application of credibility weights reflects the recognition and processing of quality differences among different data sources. Higher-quality data is assigned higher weights, thus playing a greater role in the final judgment. The weighted aggregation process is actually a key step in the fusion of multi-source information. Through reasonable weight allocation, a comprehensive assessment of the device status is achieved, avoiding erroneous judgments that may result from false alarms from a single sensor.

[0027] Step S107: When the aggregated data exceeds the real-time early warning threshold, activate the risk prediction model to generate a risk level signal and output a real-time health status report and abnormal event dataset. The activation of the risk prediction model signifies that the system has detected a potential safety hazard. At this point, the model predicts the risk's development trend and severity based on current state parameters and historical experience. The generation of risk level signals goes beyond simple over-limit alarms; it represents a risk quantification result based on deep learning, providing a scientific basis for subsequent emergency response. Real-time health status reports offer managers an intuitive overview of equipment status, while the abnormal event dataset accumulates valuable practical data for future knowledge base updates.

[0028] Step S108: Based on the real-time health status report and maintenance history, predict the probability distribution of remaining lifespan using the equipment degradation model, and output a predictive maintenance work order.

[0029] The establishment of the equipment degradation model is based on in-depth research into the evolution of equipment performance over time. Through joint analysis of historical maintenance data and condition monitoring data, typical patterns and key influencing factors of equipment performance degradation can be identified. The prediction results of the remaining life probability distribution not only provide the expected time of equipment failure, but more importantly, provide quantitative information on uncertainty, enabling maintenance decisions to be optimized under controllable risks. The generation of predictive maintenance work orders marks a fundamental shift in the entire management system from passive response to proactive prevention. By identifying maintenance needs in advance and developing targeted maintenance plans, the safe operation of equipment is ensured while reducing unnecessary maintenance costs.

[0030] In the above implementation, a complete technology chain from data acquisition to intelligent decision-making is constructed, realizing digital management and control of building equipment from design, operation, maintenance to decommissioning. The core of this solution lies in establishing a risk prediction mechanism and dynamic threshold adjustment system based on machine learning. Through multimodal sensor networks, comprehensive real-time monitoring of building equipment status is achieved, and on this basis, a closed-loop management system from anomaly detection to predictive maintenance is formed, which significantly improves the intelligence level and safety assurance capability of building equipment management.

[0031] Reference Figure 2 As one implementation of step S102, the steps of constructing a risk prediction model based on design parameters, historical accident datasets, and environmental benchmark data using machine learning algorithms, generating a dynamic threshold rule base, and outputting a benchmark digital model containing a unique equipment identifier include: Step S201: Extract features from the design parameters to generate equipment feature vectors; The core purpose of this step is to transform complex engineering drawings or product specifications into a numerical representation that computers can understand and compute, namely, equipment feature vectors. This process involves the transformation of expertise from various engineering fields. For example, the processing of "material strength indicators" typically maps them to quantitative scores based on internationally accepted standard test results and compresses them to the [0,1] interval using a normalization algorithm, thus avoiding calculation deviations caused by different units. The transformation of "load distribution parameters" may involve the application of finite element simulation technology, which abstracts the actual stress state into a stress tensor matrix and then extracts key dimensions to characterize the overall structural response capability. Furthermore, "structural connection topology," as an important parameter reflecting the interaction between components, is often described using graph theory methods to capture local weak points. After this series of feature engineering steps, the original discrete design data is reconstructed into a set of continuous and comparable high-dimensional vectors, laying a solid foundation for the next step of modeling.

[0032] Step S202: Perform accident pattern analysis on the historical accident dataset and extract the accident feature matrix; This step aims to uncover the hidden spatial distribution patterns and evolutionary trends behind past accidents. Specifically, it first requires using GIS (Geographic Information System) tools to accurately mark the locations of each accident on the corresponding 3D building model, thereby creating a risk hotspot map of the entire facility area. Then, it further extracts the geometric features of crack propagation direction, speed, and morphological evolution within these hotspot areas, organizing them into a two-dimensional or multi-dimensional array structure, the so-called accident feature matrix.

[0033] It should be noted that, to improve generalization ability, clustering algorithms are also introduced to identify common features among different types of failure modes. This ensures that the results are not only applicable to specific cases but can also be effectively transferred and applied in new contexts. This big data-driven pattern recognition strategy greatly enhances the model's sensitivity to early warning of future emergencies.

[0034] Step S203: Perform spatiotemporal correlation mapping between environmental baseline data and device feature vectors to generate an environment-device coupled feature set; Because building equipment operates in a complex and ever-changing external environment, static parameters alone cannot fully depict its true working state. Therefore, a scientifically sound spatiotemporal matching mechanism must be established to accurately map each observed climate or geological variable to the corresponding equipment unit. In practice, a sliding window method is often used to define the average or extreme values ​​within each time period, and GPS positioning information is used to cross-validate spatial coordinates. Only when the two meet a certain tolerance range (e.g., coordinate deviation less than 50 meters and time overlap exceeding 90%) can a significant correlation be confirmed and included in the joint modeling scope. The resulting "environment-equipment coupling feature set" retains the inherent attributes of each individual while incorporating external disturbance effects, providing rich sample support for deeper learning.

[0035] Step S204: Input the equipment feature vector, accident feature matrix and environment-equipment coupling feature set into the machine learning framework, and generate a risk prediction model through multimodal fusion training; "Multimodal fusion" refers to integrating different types of feature representations from multiple independent information sources within the same model architecture, allowing them to participate in the weight update and error backpropagation process. Three specialized sub-modules are employed to address their respective strengths: first, a fully connected neural network handles the structural property vectors of the device itself, its compact structure making it suitable for linear transformations; second, a convolutional neural network specializes in image-related accident feature matrices, automatically capturing visual cues such as edge textures; and finally, a time-series encoder addresses the dynamically changing environment-device coupling set, effectively capturing periodic fluctuations and abrupt changes. During the feedforward propagation stage, the outputs of the three branches are concatenated and fed into a higher-level gradient boosting decision tree model for further optimization until the optimal classification boundary is achieved. This composite modeling paradigm overcomes the limitation of traditional single-model approaches that can only perceive a single aspect, significantly improving the overall accuracy of judgment.

[0036] Step S205: Call the risk prediction model to simulate and extrapolate the environmental baseline data, and output a set of equipment degradation curves; This step is a crucial verification action carried out after training is completed. It no longer relies on real observation data but instead relies entirely on the pre-built mathematical model to conduct virtual experiments. The baseline environment scenario refers to an ideal reference condition set by humans, such as constant temperature and humidity, windless and rainless conditions, which facilitates the elimination of interfering factors and focuses on examining the aging behavior of the target object over time.

[0037] During this stage, the system repeatedly performs iterative prediction tasks on various typical devices, gradually accumulating a large amount of numerical trajectories regarding settlement rate, tilt angle, fatigue damage, and other aspects. These trajectory curves intuitively reflect the changing history of structural health, helping to detect early signs and formulate intervention measures in advance.

[0038] Step S206: Generate a dynamic threshold rule base based on the set of device degradation curves; The dynamic threshold rule base contains the mapping relationship between environmental parameters and dynamic thresholds; Specifically, the function value at a certain percentile (such as the 90th percentile) on all degradation curves is selected as the default warning line. Then, a flexible adjustment mechanism is established around this baseline. Whenever an environmental factor deviates from the normal level by a certain margin, the critical threshold is adjusted up or down proportionally. This ensures both sensitivity and robustness, truly achieving localized adaptation and real-time response.

[0039] Step S207: Generate a unique identifier for building equipment, bind and store the risk prediction model, dynamic threshold rule base and equipment unique identifier, and output the benchmark digital model.

[0040] The research findings are encapsulated into a standardized service interface for external use. A "unique identifier," similar to an ID card number, is used to uniquely distinguish each specific building facility instance. It can be automatically generated by a distributed hash algorithm, possessing global uniqueness and efficient retrieval advantages. Subsequently, a key-value database management system is used to link and save each of the aforementioned component resources to it, forming a lightweight digital copy that can be quickly loaded and restored.

[0041] The above implementation achieves effective collaboration among three heterogeneous data sources: static structural parameters of buildings, historical safety accident records, and real-time environmental variables. Through a cross-modal fusion training strategy based on spatiotemporal constraints, it effectively solves the long-standing problem of data fragmentation that has plagued the industry. It establishes a dynamic threshold determination mechanism that can automatically correct according to environmental changes, overcoming the lag and inefficiency of manual calibration.

[0042] Reference Figure 3As one implementation of step S105, the step of calling the baseline threshold in the dynamic threshold rule base and calculating the real-time early warning threshold in combination with dynamic data of the external environment includes: Step S301: Obtain the pre-generated dynamic threshold rule base and real-time monitored external environment dynamic data; The dynamic threshold rule base refers to a data structure collection constructed from multiple factors. It not only includes standard reference values ​​for various parameters under different environments but also integrates multiple variables such as building structural characteristics, geographical location differences, and equipment service life. These rules are not fixed but rather empirical threshold models derived from the analysis of a large amount of historical data. Simultaneously, the collection of dynamic external environmental data relies on a network of sensors deployed around or inside the building, such as thermometers, hygrometers, anemometers, and vibration detectors. These devices continuously provide feedback on physical signals such as temperature fluctuations, air pressure changes, and ground vibration frequencies. This dual data source constitutes the prerequisite for subsequent intelligent judgment, enabling the system to quickly retrieve standard response strategies for the corresponding situation upon sensing external disturbances.

[0043] Step S302: Analyze the categories of environmental parameters in the dynamic data of the external environment and identify the codes of key influencing factors; The system needs to classify the received raw sensor data, extract representative indicators such as temperature, humidity, and rainfall intensity, and further map them into a unified format of key influencing factor codes. These "codes" are not simple numerical symbols, but a semantic abstraction used to represent specific types of environmental events and their potential severity. For example, when rainfall exceeds a certain threshold, the system automatically assigns the identifier "RAIN_A"; if strong winds and heavy rain occur simultaneously, a higher-priority composite code "MIX_C" may be triggered. This design helps improve subsequent matching efficiency and avoids misjudgments caused by redundant parameters.

[0044] Step S303: Call the benchmark threshold that matches the key impact factor code in the dynamic threshold rule base; This step essentially demonstrates how the system utilizes existing knowledge to guide decision-making. Since the aforementioned rule base is organized and stored in a multi-dimensional matrix format, it can quickly locate the corresponding location dimension (such as the foundation location), time decay coefficient (considering equipment aging), and initially set safety boundary value based on the newly determined influence factor codes.

[0045] It should be noted that the baseline threshold is not a static constant. It is a relatively stable operational guideline with a certain degree of flexibility, derived from repeated verification in previous engineering practices. Even when facing the same type of problem scenario, the values ​​may differ due to variations in the geological structure of the region or the increase in the number of years of use. This highlights the advantage of dynamic rule bases over traditional manual experience.

[0046] Step S304: Calculate the dynamic correction coefficient based on the numerical change of the environmental parameter category; The core idea of ​​this process is to derive the proportional adjustment factor that should be applied to the original benchmark value by quantitatively assessing the deviation of the current measured results from the normal range. To ensure that the response curve is neither too sensitive nor lacks flexibility, a nonlinear function relationship is usually used for fitting and approximation, ensuring that small variations do not cause drastic oscillations, while allowing for timely amplified responses in extreme cases. In addition, considering that multiple parameters may deteriorate simultaneously under certain special operating conditions, an additional linkage compensation mechanism is needed to prevent the risk of missed reports.

[0047] Step S305: Combine the baseline threshold and the dynamic correction coefficient to synthesize the real-time warning threshold, and obtain the real-time warning threshold data packet.

[0048] For example, in hot and humid summer environments, relying on a single parameter is insufficient to accurately reflect the true level of structural safety. A specially designed synergistic effect module is necessary to fully reveal hidden risks and hazards. The resulting early warning threshold data includes not only specific alarm limits but also auxiliary information such as the location coordinates, recording timestamps, and snapshots of relevant background parameters, facilitating later traceability review and visualization.

[0049] In the above implementation, a three-dimensional rule framework covering multiple heterogeneous attributes was established, which effectively solved the problem of frequent misjudgments caused by the previous single criterion. At the same time, an intelligent coding and recognition algorithm for complex disaster scenarios was proposed, which significantly enhanced the overall robustness of the system to sudden severe weather events. In addition, the dynamic adjustment engine can autonomously complete the closed-loop management of the entire process from perception to reasoning to feedback without frequent manual intervention.

[0050] Reference Figure 4 As one implementation of step S106, the step of weighted aggregation of the data stream according to the confidence weight and comparing the aggregated data with the real-time early warning threshold includes: Step S401: Obtain the location coordinates of building equipment, the data stream of the calibration sensor network dataset, and the associated confidence weights; The reliability weighting coefficient is used to quantify the reliability of measurement results provided by different sensors or the same sensor under different spatiotemporal conditions. This weighting mechanism allows the system to evaluate the quality of each sampling point based on multiple dimensions, such as signal strength, historical stability, and external environmental factors. For example, if a temperature and humidity sensor currently reads normally, but its area has recently experienced frequent rainfall, its reliability contribution at the current moment can be reduced by a preset environmental interference compensation coefficient. Alternatively, if a sensor has reported abnormal fluctuations multiple times in the past, its stability rating will be low, thus affecting its influence ratio in the entire weighting process.

[0051] Step S402: Perform spatiotemporal window segmentation on the data stream based on the location coordinates to generate a location-related data subset; This step employs the concept of a spatiotemporal coupling model, extending the traditional data segmentation method based solely on time series to the three-dimensional spatial coordinate dimension, and further refining the time granularity on this basis.

[0052] Specifically, the system pre-constructs a spatial grid system based on the dimensions of building structural units, with each grid node representing an independent location unit. Then, based on a fixed time window (e.g., 10 seconds), the continuously flowing sensor data stream is periodically truncated, and the data is allocated to the corresponding grid partition according to the spatial location information carried by each data point.

[0053] Furthermore, to mitigate data synchronization issues caused by sensor clock drift, a 5-second time tolerance mechanism was implemented. This means that as long as the time difference between two data points does not exceed 5 seconds, they can be grouped into the same location-related subset, even if they do not belong to the exact same time window. This method effectively solves the data misalignment problem caused by hardware performance differences, improving the consistency and reliability of subsequent analysis.

[0054] Step S403: Perform weighted aggregation calculation on each location-related data subset according to the credibility weight coefficient, and output the aggregated feature value; The system employs a dynamic normalized weighted average algorithm. Its purpose is to standardize the sum of the products of each sensor reading and its corresponding weight, resulting in an aggregated feature quantity that reflects the consensus of most sensors while suppressing individual noise interference. Dynamic normalization means that the proportion of the total weight is recalculated for each aggregation operation, ensuring that even if some sensors temporarily fail or leave the network, the proportion of other valid data in the final result will not be unbalanced. More importantly, this aggregation process fully considers the potential mutual corroboration relationships and conflicts between sensors, providing a solid and accurate foundation for subsequent risk assessment.

[0055] Step S404: Call the real-time warning threshold data package and extract the real-time warning threshold that matches the location coordinates; Unlike traditional static alarm threshold settings, this solution employs a context-sensitive threshold management mechanism, binding different warning standards to specific geographical coordinates. In other words, for different functional areas within a building (such as machine rooms, stairwells, and elevator shafts), the system can configure corresponding safety boundaries based on their actual use, load-bearing characteristics, and even seasonal climate factors. This design not only meets the diverse needs of complex building environments but also enhances the adaptability and flexibility of the entire warning system. Furthermore, these thresholds can be continuously optimized and updated with accumulated operational experience, forming a self-evolving safety database.

[0056] Step S405: Compare the aggregated feature values ​​with the real-time early warning threshold in multiple dimensions to generate an abnormal state marker code; This step marks a crucial turning point in the shift from pure data analysis to intelligent diagnosis. Multi-dimensional comparison means moving beyond simply comparing single numerical values ​​to determine if a threshold has been exceeded. Instead, it introduces a series of progressive judgment criteria, forming a so-called "three-tiered progressive anomaly detection model." This model sequentially checks multiple aspects, including whether the current aggregated value deviates from the normal range, whether there is a rapid deterioration trend, and whether there is continuous exceeding of limits, corresponding to three different levels of anomaly status codes (01 Basic Anomaly, 02 Emergency Anomaly, 03 Continuous Anomaly). This helps avoid false alarms caused by accidental disturbances and also allows for timely warnings before truly significant safety hazards are encountered, improving the effectiveness and timeliness of the entire early warning and response chain.

[0057] Step S406: Output a comparison result set containing location coordinates, aggregated feature values, and abnormal state marker codes.

[0058] This comparison result set is not only an important reference for downstream implementing agencies to take countermeasures, but also first-hand information for operation and maintenance personnel to understand the on-site situation.

[0059] The above implementation overcomes the problems commonly faced by existing building monitoring systems, such as insufficient sensor redundancy, high false alarm rate, and slow response. In particular, it demonstrates excellent applicability and forward-looking advantages in real-world application scenarios where multiple heterogeneous sensors coexist, severe environmental interference, and dense and complex equipment distribution are encountered.

[0060] Reference Figure 5 As one implementation of step S107, the step of activating the risk prediction model to generate a risk level signal and outputting a real-time health status report and abnormal event dataset when the aggregated data exceeds the real-time early warning threshold includes: Step S501: Receive the comparison result set, including location coordinates, aggregated feature values, and abnormal status marker codes; Step S502: When the aggregated feature value of the abnormal state marker code exceeds the real-time warning threshold, it is determined to be an abnormal state feature, and the pre-stored risk prediction model is activated. It's important to note that the activation condition of the risk prediction model doesn't simply react immediately based on the status code of a single sampling point. Instead, it combines multiple factors, such as the anomaly type and duration, to determine whether to truly enter the deep analysis phase. For example, for false alarms caused by short-term fluctuations, the system employs a delayed judgment strategy, triggering the risk prediction model only when such anomalies occur a certain number of times or reach a certain critical state. This design effectively avoids frequent false alarms caused by sporadic noise, improving overall stability. The activated "risk prediction model" is a set of pre-trained artificial intelligence algorithm modules, typically built upon a historical failure case library. It possesses strong generalization ability and learning transfer characteristics, enabling it to quickly adapt to new scenarios and make reasonable inferences with limited new samples.

[0061] Step S503: Obtain the current device operating status parameters and environmental dynamic data stream; Among these, equipment operating status parameters encompass key variables that directly affect functional stability, such as temperature, pressure, and vibration frequency. These parameters often exhibit high-dimensional nonlinear relationships, thus requiring advanced feature engineering techniques for extraction and dimensionality reduction to facilitate model understanding and utilization. Environmental dynamic data streams refer to the changing trends of uncontrollable external factors, such as temperature fluctuations, humidity changes, and even macroscopic influencing factors like the intensity of geological activity. While these do not directly affect the target object itself, they can significantly alter the probability and speed of potential failure modes. By simultaneously accessing these two heterogeneous data sources, it is possible to ensure that the assessment more closely reflects real-world scenarios.

[0062] Step S504: Input the location coordinates, abnormal state characteristics, equipment operating status parameters and environmental dynamic data stream into the risk prediction model, perform multi-dimensional risk assessment, and output risk level code and predicted evolution path data. Specifically, unlike the traditional approach that relies solely on matching each rule from a fixed rule base, this application employs a learning-based discriminator built upon a deep neural network architecture (such as LSTM) for comprehensive judgment. This model not only captures the evolutionary patterns in time series but also uncovers hidden correlations and causal chains between different attributes, thereby achieving a higher level of cognitive understanding. Furthermore, spatial topological constraints are introduced, ensuring that each risk score is built upon a three-dimensional cognitive map, significantly improving the accuracy and interpretability of early warnings.

[0063] Risk level coding is a discrete grading standard that transforms complex uncertainties into easily communicated and managed operational instructions. This process does not simply rely on a single-dimensional indicator for a crude classification, but rather comprehensively considers the cumulative effect of multiple factors, demonstrating a certain degree of foresight and controllable adjustment capability.

[0064] Meanwhile, the output of predicted evolution path data means that the system has the ability to quantitatively characterize future development trends, which is crucial for formulating targeted intervention measures. These path descriptions are typically sets of trajectory vectors in three-dimensional space, reflecting the details of how anomalies gradually spread and proliferate over time, helping operations personnel to proactively plan and respond with resources.

[0065] Step S505: Integrate the device's unique identifier, risk level code, predicted evolution path data, and timestamp to generate a real-time health status report; All key information obtained in the early stages has been organized into a standardized document format—a real-time health status report—facilitating seamless integration and access by downstream application systems. It's worth noting that, to enhance interoperability and semantic clarity, JSON-LD, a data exchange protocol with context annotation capabilities, is used. This allows each report to carry complete metadata background information, enabling third-party platforms to accurately parse and integrate it into larger-scale urban infrastructure management systems. Furthermore, a unique identity authentication mechanism (UUID) has been incorporated to ensure that the historical records of each device are traceable and verifiable, eliminating the possibility of confusion or misinterpretation.

[0066] Step S506: Extract abnormal state features and bind location coordinates, and output the abnormal event dataset.

[0067] Specifically, by extracting a time segment covering the entire lifecycle of an event from the time window before and after each alarm action, the most representative feature vector combination is extracted and precisely mapped to the location of the incident, forming a knowledge asset that can be stored and reviewed for a long time. This not only helps technicians to deeply analyze the cause and mechanism of the accident, but also provides a valuable reference template for early identification in similar situations in the future.

[0068] The above implementation effectively solves the problems of strong lag, narrow coverage and lack of adaptive adjustment in the existing technology. Especially against the backdrop of increasing old buildings and high maintenance costs, the high flexibility and accurate prediction capabilities demonstrated by this solution will play an important role in improving the level of urban public safety.

[0069] Reference Figure 6As one implementation of step S108, the step of predicting the probability distribution of remaining lifespan and outputting predictive maintenance work orders based on real-time health status reports and maintenance history records using an equipment degradation model includes: Step S601: Obtain real-time health status report and equipment maintenance history record; This step serves as the data acquisition entry point for the entire process, aiming to build a complete equipment lifecycle profile. The "real-time health status report" refers to the status indicator data continuously collected through sensor networks deployed on various electromechanical equipment within the building (such as elevators, air conditioning units, and water pumps). This data typically covers multiple dimensions, including temperature, vibration frequency, and current fluctuations, and is analyzed through edge computing or cloud computing to form an information summary reflecting the current health level of the equipment. The "equipment maintenance history record" refers to the archival data generated from previous maintenance activities performed on this type of equipment, including but not limited to the time points of each maintenance, the specific types of operations performed, and details of replaced parts.

[0070] The core function of this stage is to provide initial data for subsequent risk assessment. On the one hand, historical experience accumulated over a long period can identify which failure modes have a high incidence rate; on the other hand, real-time monitoring can capture key signal changes that foreshadow impending anomalies, allowing for early intervention to avoid downtime losses. This combination of dynamic and static data constitutes an indispensable foundational support system for intelligent operation and maintenance.

[0071] Step S602: Analyze the risk level code and predicted evolution path data in the real-time health status report; In this stage, "risk level coding" is essentially a standardized representation used to describe the current potential failure probability of equipment. For example, equipment operating conditions are divided into three levels: low, medium, and high, and each is assigned a numerical weight coefficient (e.g., 0.3 / 0.6 / 0.9 for levels 1-3) to facilitate subsequent modeling calculations. "Predicted evolution path data" further expresses the possible trajectory of equipment performance over a future period, often relying on machine learning algorithms to extrapolate and estimate the degradation process over the next few weeks or even months.

[0072] This process requires a specially developed data decoding module to complete the semantic conversion task. Since monitoring platforms from different vendors may have different interfaces, a unified standard must be established to ensure cross-platform compatibility and portability. Furthermore, auxiliary information such as the 3D coordinates of key degradation locations needs to be extracted during the parsing process, which helps with precise positioning and visualization later. Overall, this step effectively transforms unstructured sensory data into a set of standard parameters that can be quantitatively processed.

[0073] Step S603: Extract the work order execution time, maintenance type, and parts replacement list from the maintenance history record; In this process, it is necessary to perform structural cleaning and classification on the existing maintenance history database to extract the most valuable parts for this predictive modeling. First, the "work order execution time" reflects the actual time when a certain task occurred in the past, which is crucial for tracking the service life of equipment. Second, "maintenance type" is generally divided into preventive maintenance (PM), corrective maintenance (CM), and emergency repair (EM), each representing different response strategies and the cost-benefit considerations behind them. Finally, the "parts replacement list" provides first-hand reference data on the consumption patterns of consumable parts, which is also of great significance for optimizing inventory management.

[0074] It is worth noting that reasonable filtering conditions should be set during data extraction, such as retaining only the five most recent valid service records and removing invalid or duplicate entries, to improve sample representativeness. Additionally, statistical methods should be introduced to label frequently replaced components, laying the foundation for developing targeted maintenance plans. In summary, this part of the work essentially prepares a high-quality labeled sample set for subsequent model training.

[0075] Step S604: Input the risk level code, predicted evolution path data and maintenance type into the pre-trained equipment degradation model to calculate the remaining life probability distribution; This is the most crucial and challenging step in the entire methodology—namely, how to scientifically and reasonably estimate when a specific piece of equipment will reach its critical failure threshold in the future. To address this, this invention proposes using the Weibull distribution function as the theoretical framework. This is a mathematical model widely used in reliability engineering, particularly suitable for characterizing complex mechanical equipment that exhibits a clear aging trend but lacks a constant failure rate.

[0076] Specifically, the model dynamically adjusts two important parameters: one is the shape parameter β, which is directly affected by the risk coefficient mentioned earlier. As the risk level increases, the value of β will also increase accordingly, which means that the curve becomes steeper and indicates that the failure rate increases faster. The other is the scale parameter η, which is affected by the nature of the most recent maintenance. If the last maintenance was preventive, η will be appropriately amplified, and vice versa, in order to simulate the life extension effect brought about by good maintenance.

[0077] The resulting probability density curve actually depicts a failure risk accumulation process that evolves over time. Based on this, the expected service life range under a certain confidence level can be deduced, which is the so-called "remaining life probability distribution".

[0078] Step S605: Determine the maintenance urgency level based on the remaining lifetime probability distribution; Once the future service life of the equipment is determined, the next step is to translate it into a specific action plan prioritization mechanism. A three-level classification method is used for clarity: if the predicted remaining lifespan is less than one month, it is considered a red alert (highest level), prompting immediate maintenance; those between 30 and 90 days are defined as orange alerts, recommending their inclusion in the near-term schedule; and those exceeding three months are temporarily treated as yellow observations, allowing for continued observation without immediate action.

[0079] In addition to simply considering the duration of use, the strategic importance of the equipment within the entire building operation chain is also taken into account. For example, whether it belongs to a critical power supply circuit or the main water pump set. Once it is identified as a core asset, its warning level will be automatically raised to a higher level regardless of its natural decline. The purpose of this is to prevent a chain reaction caused by neglecting secondary facilities.

[0080] Step S606: Based on the maintenance urgency level and the parts replacement list, a preliminary maintenance plan is generated by matching the preset maintenance strategy library; Once a clear priority order and necessary material reserve information are established, the process of generating customized solutions can begin. The so-called "maintenance strategy library" is essentially a pre-compiled knowledge graph-style database containing best practice templates recommended for various typical scenarios. Each strategy entry specifies applicable prerequisites, required human resource allocation, approximate time estimate, and special considerations.

[0081] To enhance flexibility and adapt to diverse real-world environments, a three-dimensional decision matrix architecture is introduced: the first dimension represents the urgency level, the second dimension represents the type of component to be repaired (mechanical, electrical, or structural), and the third dimension relates to local climate and seasonal characteristics (cold and dry winters, hot and humid summers, or high humidity during the rainy season, etc.). The cross-combination of these three variables quickly identifies the most suitable comprehensive contingency plan for the current situation, significantly improving response efficiency and consistency.

[0082] Step S607: Obtain the building topology and optimize the spatial execution path of the preliminary maintenance plan based on the location coordinates; Considering the characteristics of modern high-rise buildings, such as intricate internal pipelines and narrow passages, even a detailed technical plan may not be able to be successfully implemented on-site. Therefore, the concept of Geographic Information System (GIS) can be introduced, and the rich geometric attributes and connections provided by Building Information Modeling (BIM) can be used to further refine the spatial layout of the entire operation process.

[0083] Specifically, the exact installation location of the target equipment is first precisely located. Then, the optimal movement route is planned around this center point to the tool storage area, backup power access point, and even the nearest safe evacuation exit. During this process, the work area occupied by other ongoing construction projects must be fully considered, and avoidance actions should be taken as needed to prevent mutual interference. The final path not only includes clearly visible 3D navigation guidance but also the estimated travel time and potential safety hazard assessment for each turning point along the way, greatly enhancing the operational controllability for on-site personnel.

[0084] Step S608: Integrate the preliminary maintenance plan and spatial execution path, and output predictive maintenance work orders.

[0085] Among these measures, all interim results are compiled and packaged into a standard electronic document format that can be directly accessed by frontline employees, namely, a predictive maintenance work order.

[0086] In one embodiment of this application, the document should include at least the following parts: first, the equipment's identification code for easy traceability; second, the selected specific maintenance package number and related process procedure links; third, the detailed specifications and models of the required replacement materials and their current inventory status; fourth, the reference address of the aforementioned spatial navigation file; and fifth, a series of standardized safety operation manual chapter indexes.

[0087] The above implementation creates a highly intelligent, self-evolving integrated operation and maintenance service platform. Compared to the traditional passive management model that relies on manual inspections and post-event repairs, this technical solution significantly improves equipment availability, reduces operating costs, and decreases the probability of unexpected accidents. It is particularly suitable for critical infrastructure support in densely populated areas such as large commercial complexes, hospitals, and schools.

[0088] Reference Figure 7 As a further implementation of the building equipment lifecycle management method, the following steps are included after generating predictive maintenance work orders: Step S701: Receive the record of executed predictive maintenance work orders; The maintenance work order records include, but are not limited to, the actual tool model used, the operator's number, the work time, and the corresponding measured values ​​of process parameters (such as bolt torque, welding current, etc.). This information differs from the ideal settings in the early planning stage; it reflects the immediate adjustments brought about by changes in on-site conditions, and is therefore more representative and reliable.

[0089] Step S702: Deconstruct the actual maintenance operation list in the executed predictive maintenance work order and extract the component replacement code and process parameters; Among them, the component replacement coding adopts the internationally recognized classification standard. For example, vague expressions such as "replacing the spindle bearing" are converted into machine-recognizable code identifiers such as "MD-2154B" to facilitate subsequent retrieval and statistical analysis. The extraction of process parameters not only involves the identification of numerical variables (such as laser calibration accuracy, hydraulic pressure, hot melt adhesive curing time, etc.), but also requires the use of natural language processing technology to parse out quality evaluation statements with subjective judgment, and then convert them into rating levels with certain semantic meaning.

[0090] Step S703: Collect equipment operation data sets within a preset period before and after maintenance, calculate and compare key energy efficiency ratio parameters before and after maintenance, and generate maintenance quality assessment scores. The system includes a comparative analysis process for equipment operation before and after maintenance. The preset cycle should cover the entire evolution process from the initial unstable commissioning phase to stable operation and the onset of potential degradation. All real-time sensor data collected during this period will be used to calculate key energy efficiency ratio parameters.

[0091] In this embodiment, an observation window covering the entire lifecycle fluctuation cycle is selected, typically set to one month in length, to comprehensively capture the short-term adaptive response and long-term performance caused by newly installed components. By calculating key indicators such as energy efficiency during this period and introducing environmental compensation factors to mitigate external interference, more accurate and reliable energy efficiency evaluation results are obtained. It should be noted that an asymmetric weighted sampling principle is adopted in the sample selection strategy, giving lower weight to data that has recently undergone maintenance and higher attention to data at the end of a stable operation phase, thereby highlighting the focus on long-term stability.

[0092] Furthermore, instead of using the traditional average comparison method, a time-decay weighted difference approach was introduced to determine the differences between recent observations, giving them greater influence. Considering the significant differences in difficulty among different types of maintenance tasks, a process complexity adjustment mechanism was also embedded in the score generation process. This mechanism automatically matches the appropriate multiplication factor based on a pre-established process library, making the final evaluation score more fair and representative.

[0093] Step S704: Adjust the remaining life calculation parameters of the equipment degradation model based on the maintenance quality assessment score; Equipment degradation models refer to a probabilistic statistical framework used to predict the likelihood of equipment failure at a future point in time, often constructed using forms such as the Weibull distribution. Furthermore, each new maintenance practice requires targeted optimization and upgrades to provide feedback.

[0094] Specifically, based on the maintenance quality score, several important parameters in the original model are appropriately modified according to preset rules. If the maintenance is successful, the overall lifespan estimated by the model can be appropriately increased; conversely, if potential problems are found, the corresponding early warning mechanism should be activated to accelerate the aging process simulation.

[0095] Step S705: Recalculate the probability distribution of remaining equipment life based on the corrected equipment degradation model, and update the risk prediction model parameters in the baseline digital model.

[0096] Based on this revised model, the system recalculates the probability distribution of the device's remaining useful life (RUL), which not only includes the possible future failure time of the device, but also quantifies the corresponding level of uncertainty.

[0097] Subsequently, this updated remaining lifespan probability distribution information was directly used to optimize the risk prediction module in the baseline digital model. The updates to the risk prediction model parameters are mainly reflected in two dimensions: first, adjusting the time window and threshold settings for risk assessment to match the current health status of the equipment; second, optimizing the calculation weights and correlation coefficients of risk probabilities to ensure the accuracy and timeliness of risk warnings.

[0098] In the above implementation, this dynamic update mechanism ensures that the entire digital model is always based on the latest equipment status information for risk prediction, effectively improving prediction accuracy and providing reliable data support for making accurate maintenance decisions, ultimately realizing the transformation from passive maintenance to proactive prevention.

[0099] This application also discloses a building equipment full lifecycle management system.

[0100] A building equipment lifecycle management system, the management system includes: The acquisition module is used to acquire the design parameters of building equipment, historical accident datasets, and environmental baseline data; The model building module is used to build a risk prediction model based on design parameters, historical accident datasets and environmental benchmark data, and generate a dynamic threshold rule base, outputting a benchmark digital model containing the equipment's unique identifier. The sensor configuration module is used to deploy a multimodal sensor group on the physical structure of building equipment based on a benchmark digital model, bind each sensor to a unique device identifier, perform noise filtering and confidence weight assignment on the initial data collected by the multimodal sensor group, and output a calibration sensor network dataset. The acquisition module is used to acquire the data stream of the calibration sensor network dataset and dynamic data of the external environment in real time; The early warning threshold calculation module is used to call the baseline threshold in the dynamic threshold rule library and combine it with dynamic data of the external environment to calculate the real-time early warning threshold; The aggregation and comparison module is used to weight and aggregate the data stream according to the confidence weight, and compare the aggregated data with the real-time warning threshold. The risk prediction module is used to activate the risk prediction model to generate risk level signals and output real-time health status reports and abnormal event datasets when the aggregated data exceeds the real-time warning threshold. The maintenance work order generation module is used to predict the probability distribution of remaining lifespan based on real-time health status reports and maintenance history records, and output predictive maintenance work orders through equipment degradation models.

[0101] The building equipment lifecycle management system of this application embodiment can implement any of the above methods, and the specific working process of each module in the system can refer to the corresponding process in the above method embodiment.

[0102] In the several embodiments provided in this application, it should be understood that the provided methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for example, the division of a certain module is merely a logical functional division, and in actual implementation there may be other division methods, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0103] This application also discloses a computer device.

[0104] Computer equipment includes memory, processor, and computer program stored in memory and executable on the processor. When the processor executes the computer program, it implements a building equipment lifecycle management method as described above.

[0105] This application also discloses a computer-readable storage medium.

[0106] A computer-readable storage medium storing a computer program that can be loaded by a processor and executed as described above in any of the building equipment lifecycle management methods.

[0107] The computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device; the program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0108] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A building equipment whole life cycle management method, characterized by, The management method includes: Obtain building equipment design parameters, historical accident datasets, and environmental baseline data; Based on the design parameters, historical accident datasets, and environmental benchmark data, a risk prediction model is constructed using machine learning algorithms, and a dynamic threshold rule base is generated, outputting a benchmark digital model containing the unique identifier of the equipment. Based on the aforementioned benchmark digital model, a multimodal sensor group is deployed on the physical structure of the building equipment. Each sensor is bound to the unique identifier of the equipment, and noise filtering and confidence weight assignment are performed on the initial data collected by the multimodal sensor group to output a calibration sensor network dataset. Real-time acquisition of the data stream of the calibration sensor network dataset and dynamic data of the external environment; The baseline threshold in the dynamic threshold rule base is called, and the real-time early warning threshold is calculated by combining the dynamic data of the external environment. The data stream is weighted and aggregated according to the credibility weight, and the aggregated data is compared with the real-time warning threshold. When the aggregated data exceeds the real-time warning threshold, the risk prediction model is activated to generate a risk level signal and output a real-time health status report and abnormal event dataset. Based on the real-time health status report and maintenance history, the remaining lifespan probability distribution is predicted through the equipment degradation model, and predictive maintenance work orders are output.

2. The building equipment lifecycle management method of claim 1, wherein, Based on the aforementioned design parameters, historical accident datasets, and environmental benchmark data, the steps of constructing a risk prediction model using machine learning algorithms, generating a dynamic threshold rule base, and outputting a benchmark digital model containing unique equipment identifiers include: Feature extraction is performed on the design parameters to generate a device feature vector; Accident pattern analysis is performed on the historical accident dataset to extract the accident feature matrix; The environmental baseline data and the device feature vectors are spatiotemporally correlated and mapped to generate an environment-device coupled feature set. The equipment feature vector, accident feature matrix, and environment-equipment coupled feature set are input into a machine learning framework, and a risk prediction model is generated through multimodal fusion training. The risk prediction model is invoked to simulate and extrapolate the environmental baseline data, and a set of equipment degradation curves is output. A dynamic threshold rule library is generated based on the aforementioned set of device degradation curves; Generate a unique identifier for the building equipment, bind and store the risk prediction model, dynamic threshold rule base and the unique identifier, and output a benchmark digital model.

3. The building equipment lifecycle management method of claim 1, wherein, The steps of calling the baseline threshold in the dynamic threshold rule base and calculating the real-time early warning threshold in combination with the dynamic data of the external environment include: Acquire pre-generated dynamic threshold rule base and real-time monitored external environmental dynamic data; Analyze the categories of environmental parameters in the dynamic data of the external environment and identify the coding of key influencing factors; Call the benchmark threshold in the dynamic threshold rule base that matches the key impact factor code; Calculate the dynamic correction coefficient based on the numerical change of the environmental parameter categories; By combining the baseline threshold with the dynamic correction coefficient, a real-time early warning threshold is synthesized, resulting in a real-time early warning threshold data packet.

4. The method for full lifecycle management of building equipment according to claim 3, characterized in that, The step of weighted aggregation of the data stream based on the credibility weight and comparing the aggregated data with the real-time early warning threshold includes: Obtain the location coordinates of building equipment, the data stream of the calibration sensor network dataset, and the associated confidence weights; The data stream is segmented into a spatiotemporal window based on the location coordinates to generate a location-related data subset. Based on the credibility weight coefficient, a weighted aggregation calculation is performed on each location-related data subset, and the aggregated feature value is output. The real-time warning threshold data packet is invoked to extract the real-time warning threshold that matches the location coordinates; The aggregated feature values ​​are compared with the real-time early warning threshold in multiple dimensions to generate an abnormal state marker code; The output includes a set of comparison results containing the location coordinates, aggregated feature values, and abnormal state marker codes.

5. The method for full lifecycle management of building equipment according to claim 4, characterized in that, When the aggregated data exceeds the real-time early warning threshold, the steps to activate the risk prediction model to generate a risk level signal and output a real-time health status report and abnormal event dataset include: Receive the comparison result set, including location coordinates, aggregated feature values, and abnormal status marker codes; When the aggregated feature value of the abnormal state flag code exceeds the real-time warning threshold, it is determined to be an abnormal state feature, and the pre-stored risk prediction model is activated. Obtain current device operating status parameters and dynamic environmental data stream; The location coordinates, abnormal state characteristics, equipment operating status parameters and environmental dynamic data stream are input into the risk prediction model to perform multi-dimensional risk assessment and output risk level codes and predicted evolution path data. Integrate the device's unique identifier, risk level code, predicted evolution path data, and timestamp to generate a real-time health status report; Extract abnormal state features and bind them to location coordinates to output an abnormal event dataset.

6. The method for full lifecycle management of building equipment according to claim 5, characterized in that, Based on the real-time health status report and maintenance history, the steps for predicting the remaining lifespan probability distribution using the equipment degradation model and outputting predictive maintenance work orders include: Obtain real-time health status reports and equipment maintenance history records; Analyze the risk level coding and predicted evolution path data in the real-time health status report; Extract the work order execution time, maintenance type, and component replacement list from the maintenance history record; The risk level code, predicted evolution path data, and maintenance type are input into a pre-trained equipment degradation model to calculate the remaining lifetime probability distribution. The maintenance urgency level is determined based on the remaining lifetime probability distribution. Based on the maintenance urgency level and the component replacement list, a preliminary maintenance plan is generated by matching the preset maintenance strategy library; Obtain the building topology and optimize the spatial execution path of the preliminary maintenance plan based on the location coordinates; Integrate the preliminary maintenance plan and spatial execution path to output predictive maintenance work orders.

7. A method for full lifecycle management of building equipment according to any one of claims 1 to 6, characterized in that, After generating predictive maintenance work orders, the following steps are also included: Receive records of executed predictive maintenance work orders; Deconstruct the actual maintenance operation list in the executed predictive maintenance work order, and extract the component replacement codes and process parameters; Collect equipment operation datasets within a preset period before and after maintenance, calculate and compare key energy efficiency ratio parameters before and after maintenance, and generate maintenance quality assessment scores. The remaining life calculation parameters of the equipment degradation model are adjusted based on the maintenance quality assessment score. Based on the revised equipment degradation model, the probability distribution of remaining equipment life is recalculated, and the risk prediction model parameters in the baseline digital model are updated.

8. A building equipment full lifecycle management system, characterized in that, The management system includes: The acquisition module is used to acquire the design parameters of building equipment, historical accident datasets, and environmental baseline data; The model building module is used to build a risk prediction model based on the design parameters, historical accident dataset and environmental benchmark data, through machine learning algorithms, generate a dynamic threshold rule base, and output a benchmark digital model containing the unique identifier of the equipment. The sensor configuration module is used to deploy a multimodal sensor group on the physical structure of building equipment based on the benchmark digital model, bind each sensor to the unique identifier of the equipment, perform noise filtering and confidence weight assignment on the initial acquisition data of the multimodal sensor group, and output a calibration sensor network dataset. The acquisition module is used to acquire the data stream of the calibration sensor network dataset and the dynamic data of the external environment in real time; The warning threshold calculation module is used to call the baseline threshold in the dynamic threshold rule library and combine it with the dynamic data of the external environment to calculate the real-time warning threshold. The aggregation and comparison module is used to perform weighted aggregation on the data stream according to the credibility weight, and compare the aggregated data with the real-time warning threshold. The risk prediction module is used to activate the risk prediction model to generate a risk level signal and output a real-time health status report and abnormal event dataset when the aggregated data exceeds the real-time warning threshold. The maintenance work order generation module is used to predict the probability distribution of remaining lifespan based on the real-time health status report and maintenance history, and output predictive maintenance work orders.

9. A computer device, characterized in that: The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer program is stored that can be loaded by a processor and executed as described in any one of claims 1 to 7.

Images