Electromechanical system health monitoring and optimizing method based on artificial intelligence

Abstract

The invention belongs to the technical field of electromechanical equipment intelligent monitoring and operation and maintenance management, and discloses an electromechanical system health monitoring and optimization method based on artificial intelligence, and the method comprises the steps: building a digital twin architecture based on a simulation model, reasonably arranging a multi-source sensor, and carrying out the optimization of the health of an electromechanical system; key equipment is fully covered; the system acquires operation data at high frequency and synchronously transmits the operation data to the cloud through multiple channels to realize real-time monitoring; the collected data is cleaned and preprocessed to remove abnormities so as to ensure the data quality; inputting the cleaned data into a neural network, extracting features, and identifying an operation trend and a fault mode; establishing a prediction model in combination with historical data, and constructing a maintenance knowledge base; the equipment state is predicted in real time, and early warning is dynamically generated; through virtual-real synchronization, health and carbon emission states are mapped in real time, equipment operation is intelligently optimized, and cloud closed-loop monitoring is continuously iterated.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent monitoring and operation and maintenance management of electromechanical equipment, and specifically provides an electromechanical system health monitoring and optimization method based on artificial intelligence. Background Art

[0002] With the in-depth penetration of smart city construction and green and low-carbon concepts, the operation and management of infrastructure electromechanical systems are facing a complex dilemma. The traditional operation and maintenance model focuses on structural safety, but pays insufficient attention to real-time monitoring and preventive maintenance of the health status of electromechanical systems. This "structural emphasis, electromechanical neglect" mentality leads to frequent sudden failures of equipment during operation. The operation and maintenance team often responds passively after the failure occurs, which not only increases maintenance costs but may also affect the normal operation of infrastructure.

[0003] The existing monitoring system also has obvious shortcomings. It mainly collects single parameters and cannot fully perceive the multi-dimensional conditions of equipment vibration, temperature, electrical parameters, fluid parameters, etc. There is a lack of effective linkage between the monitoring modules, and the data forms information islands, which greatly reduces the monitoring accuracy and makes it difficult to accurately predict potential equipment failures. In addition, the current carbon emission management only focuses on the operation stage of the electromechanical system, ignoring the carbon emissions of the equipment throughout its life cycle from production, transportation, installation to scrapping. Under this management model, it is impossible to dynamically track and optimize the carbon emissions of the equipment throughout its life cycle. Summary of the Invention

[0004] The purpose of the present invention is to provide an artificial intelligence-based electromechanical system health monitoring and optimization method to solve the problems raised in the above background technology.

[0005] To achieve the above objectives, the present invention provides the following technical solutions: an artificial intelligence-based electromechanical system health monitoring and optimization method, which builds a digital twin architecture based on a simulation model, rationally deploys multi-source sensors, and comprehensively covers key equipment; the system collects operating data at high frequency and transmits it to the cloud synchronously through multiple channels to achieve real-time monitoring; the collected data is cleaned and preprocessed to remove anomalies and ensure data quality; the cleaned data is input into a neural network to extract features and identify operating trends and failure modes; a prediction model is established based on historical data to build a maintenance knowledge base; equipment status is predicted in real time and early warnings are dynamically generated; through virtual-real synchronization, health and carbon emission status are mapped in real time, equipment operation is intelligently optimized, and cloud-based closed-loop monitoring is continuously iterated; the specific steps of the optimization method are as follows: S1: Establishing a multi-source sensor network and determining equipment monitoring targets: Based on the finite element analysis model, transient simulation model, and carbon emission calculation model of the building and electromechanical systems, a digital twin infrastructure for the electromechanical system is established. Through model simulation and operating condition analysis, key monitoring targets and representative parts prone to failure are selected, and monitoring points are determined. Targeted at different equipment and environmental parameters, NB-IoT wireless vibration sensors (sampling frequency not less than 500Hz), thermal imaging sensors (resolution not less than 160×120 pixels, refresh rate ≥9Hz), current and voltage sensors (data refresh cycle ≤1s), flow and pressure sensors (response time ≤0.5s), and environmental sensors are deployed. The number of measurement points is configured to be no less than 3 sets of sensor units per 100㎡ of building area to ensure comprehensive data coverage; S2: Real-time data high-frequency acquisition and multi-channel wireless transmission: The collected multi-source monitoring data is asynchronously transmitted in real time via NB-IoT (latency ≤ 2s), 5G (latency ≤ 0.5s), and the RS485 bus (data transmission rate ≥ 115.2kbps). Based on a transmission architecture that combines wireless and wired networks, it collects multiple parameters such as vibration, temperature, current, voltage, flow, pressure, humidity, and illumination during the operation of the electromechanical system at high frequency and in real time. The multi-channel data is then synchronously transmitted to the cloud or edge computing platform, ensuring data continuity, stability, and high timeliness. S3: Health monitoring raw data cleaning and outlier preprocessing: For multi-dimensional data collected in real time, we use data processing techniques such as sliding window analysis, time-domain averaging, and wavelet denoising to complete outlier removal, data completion, and standardization. This ensures the integrity, accuracy, and temporal continuity of health monitoring data and eliminates interference from environmental noise and load fluctuations on subsequent analysis. S4: Electromechanical System Operation Status Feature Extraction and Multi-Scale Trend Analysis: A three-layer convolutional neural network (convolution kernel size 3×3, stride 1, and output channels of 32, 64, and 128, respectively) is used to perform deep feature extraction on four parameters: vibration, thermal, electrical, and fluid. Peak value, RMS value, kurtosis, and spectral peak displacement are extracted from time-domain vibration signals. Total harmonic distortion (THD) and current surge characteristics are extracted from electrical parameters. The maximum temperature difference in high-temperature zones is extracted from thermal imaging. Flow rate fluctuation is extracted from fluid parameters. Equipment operation trends are calculated using time-series data streams to identify potential failure modes. S5: Historical Condition Modeling and Predictive Maintenance Knowledge Base Construction: Based on collected historical equipment condition data, a knowledge base for electromechanical equipment condition monitoring is established. Data-driven support vector machine (SVM) and long short-term memory (LSTM) prediction models are constructed. Input features include: root mean square vibration, harmonic distortion, thermal imaging hot spot temperature, and flow rate fluctuation. The training set must contain ≥ 100,000 data sets, with a prediction error target of less than 5%. The remaining service life and health index of the equipment are output. A knowledge base for equipment failure cases and an operation and maintenance response strategy library are also constructed simultaneously. S6: Real-time status prediction and graded dynamic warning mechanism: Utilizes an LSTM model to predict the equipment's operating status over the next 24 hours, focusing on typical fault signs such as sudden vibration spikes, current distortion, and abnormal temperature rise in thermal imaging. Three levels of dynamic warning thresholds are set: mild anomaly (prediction deviation 5%-10%), moderate anomaly (prediction deviation 10%-15%), and severe anomaly (prediction deviation ≥15%). The warning generation time lag is controlled to ≤1 second, and the system supports alarm linkage and on-site inspection dispatching. S7: Digital Twin Virtual-Real Synchronization and Full Lifecycle Carbon Emissions Mapping: Through the bidirectional integration of finite element simulation models and real-time monitoring data, a digital twin platform for electromechanical systems is established, ensuring virtual-real data synchronization delays of ≤2 seconds and virtual-real residuals of ≤5%. Based on the real-time load, energy efficiency coefficient, and operating time of the equipment, the carbon emissions of each device are dynamically calculated, and a cumulative carbon emissions curve for the full lifecycle of the electromechanical system is generated. The equipment health index and carbon emission forecast results are simultaneously updated; S8: Intelligent optimization scheduling and priority operation strategy implementation: Based on the equipment health index, energy efficiency parameters and carbon emission indicators of the digital twin platform, the equipment start-up and shutdown sequence, load distribution and operation priority are dynamically adjusted. Equipment with good health and higher energy efficiency is prioritized to achieve optimal overall system operation efficiency and minimize carbon emissions, avoiding continuous fatigue of high-load equipment. S9: Cloud-based remote O&M monitoring and closed-loop feedback optimization: Real-time display of equipment health status, early warning information, carbon emissions, and scheduling optimization results is provided via the web, app, or local control terminal. Cloud-based O&M decision-making is synchronized with on-site control systems to form an intelligent O&M closed-loop management system. Integrating new data from subsequent equipment operations, it continuously iterates and optimizes predictive models and scheduling strategies, enhancing the health management capabilities of electromechanical systems throughout their entire lifecycle.

[0006] Preferably, the specific steps of establishing a multi-source sensor network and determining equipment monitoring objects in S1 are as follows: Step 1: Digital Twin Architecture Construction and Monitoring Planning: Relying on finite element analysis, transient simulation, and carbon emission calculation models of buildings and electromechanical systems, we build a digital twin infrastructure for electromechanical systems. Through in-depth simulation and operating condition analysis of the model, we accurately screen key monitoring objects and fault-prone areas, and scientifically determine monitoring points. This provides a reliable basis for real-time perception and precise control of the subsequent electromechanical system operating status, laying a solid foundation for digital twin applications. Step 2: Multi-source sensor deployment and measurement point optimization: Based on equipment characteristics and environmental parameter differences, rationally deploy multiple types of sensors, such as NB-IoT wireless vibration, thermal imaging, current and voltage, and clearly define the performance parameter requirements for each sensor. At the same time, adhere to the principle of deploying no fewer than three sensor units per 100 square meters of building area. Combined with the multi-source sensor measurement point optimization logic, the number of measurement points is dynamically planned taking into account factors such as building area to comprehensively ensure the integrity and effectiveness of monitoring data coverage. The optimization expression formula of multi-source sensor measurement points is: ; Where, Minimum number of sensors to be deployed (pieces) Monitoring area (㎡) Ceiling function; Preferably, the specific steps of high-frequency real-time data acquisition and multi-channel wireless transmission in S2 are as follows: Step 1: Build a multi-channel hybrid transmission architecture: Adopt an asynchronous multi-channel transmission mode that combines NB-IoT, 5G, and the RS485 bus to establish a complementary wireless and wired data transmission architecture. NB-IoT ensures wireless transmission efficiency with a latency of no more than 2s, and 5G with a low latency of ≤0.5s. The RS485 bus ensures stable wired transmission with a data transmission rate of ≥115.2kbps. The three transmission modes work together to build a solid and efficient channel foundation for electromechanical system data transmission. Step 2: Multi-parameter high-frequency acquisition and data synchronization: Based on the aforementioned transmission architecture, high-frequency real-time acquisition of multiple parameters, such as vibration and temperature, is performed during electromechanical system operation. Using multi-channel data synchronization logic, the various types of collected data are quickly and stably transmitted synchronously to the cloud or edge computing platform via different transmission channels. This logic fully considers the characteristics of each transmission channel to ensure high-frequency data acquisition and continuous, stable, and timely transmission, providing reliable data support for system analysis and decision-making. The expression formula for multi-channel data synchronization is: ; Where, time Synchronous data collection at all times, No. Data after delayed synchronization of channel sensors, No. Channel transmission delay (seconds), The total number of sensor channels. Multi-channel synchronous data needs to be delayed compensated before the timestamp can be unified.

[0007] Preferably, the specific steps of health monitoring raw data cleaning and outlier preprocessing in S3 are as follows: Step 1: Application of multidimensional data processing technology: For the multidimensional data of electromechanical systems collected in real time, a combination of sliding window analysis, time domain averaging, and wavelet denoising is used to process data. Sliding window analysis dynamically captures data features, time domain averaging is used to integrate data information, and wavelet denoising is used to filter out interference components in the data. In this process, outliers are accurately identified and eliminated based on abnormal data elimination logic. Data completion and standardization are also performed simultaneously to comprehensively optimize the original data. Step 2: Data Quality Assurance and Interference Elimination: Through a series of data processing operations, the integrity, accuracy, and temporal continuity of health monitoring data are effectively ensured. The application of abnormal data elimination logic avoids the impact of outliers on the overall data. Data completion and standardization processing improves data content and unifies the format. These processes eliminate interference from factors such as environmental noise and load fluctuations on subsequent data analysis, providing a reliable data foundation for data-based electromechanical system status assessment and fault diagnosis. The expression formula for abnormal data removal is: ; Where, Standard deviation score (Z-score), Collect data, The mean value in the sliding window, Standard deviation within the sliding window, abnormal data judgment: When |Z|>3, the data is judged to be abnormal and needs to be eliminated.

[0008] Preferably, the specific steps of extracting the operating status features of the electromechanical system and performing multi-scale trend analysis in S4 are as follows: Step 1: Multi-parameter deep feature extraction: Using a three-layer convolutional neural network with a 3×3 kernel, a stride of 1, and output channels of 32, 64, and 128, we perform deep feature mining on four parameters: vibration, thermal, electrical, and fluid. For time-domain vibration signals, we extract peak values, root mean square (RMS), and other features using specific calculation logic. For electrical parameters, we focus on total harmonic distortion and current impact. For thermal imaging, we focus on the maximum temperature difference in high-temperature areas. For fluid parameters, we extract flow rate fluctuations, achieving feature integration of multi-source data. Step 2: Equipment Failure Pattern Identification: This process converts the extracted multi-parameter features into a time-series data stream. Based on this data stream, the system analyzes equipment operating trends. By continuously monitoring and analyzing various parameter features and combining key indicator changes in the root mean square (RMS) of the vibration signal, the system accurately identifies potential equipment failure modes and predicts equipment operating status in advance, providing strong support for electromechanical system maintenance and fault warning. The expression formula of the root mean square of the vibration signal is: ; Where, RMS: root mean square value of vibration signal, No. The vibration acceleration of each sampling point, The total number of sampling points, the root mean square of vibration is one of the key indicators of equipment health status, and is used to monitor the vibration level of the equipment in real time.

[0009] Preferably, the specific steps of historical state modeling and predictive maintenance knowledge base construction in S5 are as follows: Step 1: Prediction Model Construction and Training: Based on the collected historical equipment status data, a knowledge base for electromechanical equipment condition monitoring is established. Using support vector machines and long short-term memory (LSTM) networks, with vibration root mean square (RMS) and harmonic distortion rate features as input, a data-driven prediction model is constructed. Deep training is performed using no less than 100,000 training data sets. The model is optimized using LSTM's state prediction logic, strictly controlling the prediction error to within 5%, and accurately outputting the equipment's remaining service life and health index. Step 2: Collaboratively build a knowledge base system: While completing the prediction model construction, simultaneously build an equipment failure case knowledge base and an operation and maintenance response strategy library. The failure case knowledge base summarizes historical failure information and provides a reference for problem diagnosis; the operation and maintenance response strategy library formulates targeted treatment plans, achieving full-process knowledge collaboration from fault prediction and diagnosis to disposal, and improving the intelligent operation and maintenance system of electromechanical equipment. The expression formula of LSTM state prediction is: ; Where, The current hidden state (prediction output), The hidden state at the previous moment, Input features at the current moment (including vibration, electrical parameters, thermal imaging, etc.), weight matrix, Bias term; single-step prediction structure based on long short-term memory network, applied to future state prediction.

[0010] Preferably, the specific steps of the real-time status prediction and hierarchical dynamic early warning mechanism in S6 are as follows: Step 1: LSTM model trend prediction and fault monitoring: The LSTM model is used to predict the equipment's operating status over the next 24 hours, focusing on capturing typical fault symptoms such as sudden increases in vibration, current distortion, and abnormal temperature rise in thermal imaging. The LSTM state prediction logic processes time series data to accurately predict equipment operating trends, providing a reliable basis for fault warning. Step 2: Three-level dynamic warning mechanism and system functions: A three-level dynamic warning threshold system is established: mild anomaly (prediction deviation 5%-10%), moderate anomaly (10%-15%), and severe anomaly (≥15%). The warning generation lag time is controlled within 1 second. This mechanism, combined with dynamic threshold logic, automatically triggers the corresponding warning level based on the prediction deviation. The system also supports alarm linkage and on-site inspection dispatch functions, forming a closed-loop management from prediction to disposal. The expression formula of the three-level dynamic warning threshold is: ; Where, Percentage of prediction error, Predict state values ​​(e.g. predict vibration), Current actual monitoring status value; Warning level: Mild abnormality: ; Moderate abnormality: ; Severe abnormalities: , dynamically divide the warning level based on the prediction error and promptly prompt the change of equipment status.

[0011] Preferably, the specific steps of digital twin virtual-real synchronization and full life cycle carbon emission mapping in S7 are as follows: Step 1: Build a digital twin platform for the electromechanical system: Through the bidirectional integration of finite element simulation models and real-time monitoring data, a digital twin platform for the electromechanical system is constructed. Relying on virtual-real residual control logic, the quality of data synchronization is strictly controlled to ensure that the virtual-real data synchronization delay is ≤ 2 seconds and the virtual-real residual is ≤ 5%, achieving accurate mapping and dynamic linkage between the physical system and the virtual model. Step 2: Dynamic Management of Carbon Emissions and Health Index: Based on the real-time load, energy efficiency coefficient, and operating hours of the equipment, the carbon emissions of each device are dynamically calculated, and a cumulative carbon emission curve for the entire life cycle of the electromechanical system is generated. At the same time, the platform updates the equipment health index and carbon emission forecast results in real time, realizing the coordinated monitoring and management of equipment operating status and environmental indicators. The digital twin virtual-real residual control expression is: ; Where, The percentage of virtual and real residuals, Digital twin virtual state data, Real-time actual monitoring data.

[0012] Preferably, the specific steps for implementing the intelligent optimization scheduling and priority operation strategy in S8 are as follows: Step 1: Dynamic Scheduling Strategy Development: Develop a dynamic scheduling plan based on the digital twin platform's device health index, energy efficiency parameters, and carbon emission indicators. Based on the health-first scheduling priority logic, determine the device start-up and shutdown sequence, load distribution, and operation priority. Prioritize scheduling devices with the best health and high energy efficiency, laying the foundation for efficient system operation. Step 2: Achieve system operation optimization goals: Through the above-mentioned dynamic scheduling, optimize the overall system operation efficiency and minimize carbon emissions. Reasonable load distribution avoids continuous fatigue of high-load equipment and extends equipment life. At the same time, while ensuring stable system operation, balance efficiency and environmental protection to achieve comprehensive benefits. The health-first scheduling priority expression formula is: ; Where, Equipment operation priority score, Equipment health index (0-1), The current energy efficiency coefficient of the equipment (0-1), The current carbon emissions per unit time of the equipment ( ).

[0013] Preferably, the specific steps of cloud-based remote operation and maintenance monitoring and closed-loop feedback optimization in S9 are as follows: Step 1: Multi-terminal information display and O&M closed-loop: Through the web, app, and local control terminals, equipment health status, early warning information, carbon emission data, and scheduling optimization results are presented in real time. Cloud-based O&M decision-making and on-site control systems are synchronized, and the time control logic of cloud-based alarm responses is followed to form an intelligent O&M closed-loop management to ensure efficient O&M connection. Step 2: Model iteration and management capability enhancement: Integrate new equipment operation data to continuously iterate and optimize prediction models and scheduling strategies. By continuously improving model accuracy and strategy adaptability, the health management capabilities of electromechanical systems throughout their lifecycle will be gradually enhanced, achieving efficient management of the entire process from real-time monitoring and intelligent decision-making to continuous optimization. The cloud alarm response delay expression formula is: ; Where, Total alarm response delay (seconds), Data upload transmission delay (seconds), Cloud processing computing latency (seconds), The cloud-to-client alarm push delay (seconds) ensures that Ttotal ≤ 3 seconds to meet the cloud's real-time alarm response requirements.

[0014] The beneficial effects of the present invention are as follows: 1. By constructing a multi-dimensional state feature extraction and predictive maintenance algorithm based on artificial intelligence, the present invention can collect and conduct deep learning analysis on key data such as vibration, electrical parameters, and thermal imaging of the electromechanical system in actual operation in real time. Through the joint application of convolutional neural networks and long-short-term memory networks, the equipment operation trends and potential fault characteristics are extracted to achieve accurate prediction of equipment status. Compared with the traditional model of relying on periodic maintenance or post-maintenance in the existing technology, the present invention can significantly identify equipment anomalies in advance, effectively reduce the probability of sudden failures, and reduce the need for emergency repairs, thereby improving the continuity and stability of system operation, extending the service life of equipment, and comprehensively optimizing the operation and maintenance efficiency of electromechanical systems.

[0015] 2. The present invention breaks through the single data acquisition mode of the existing monitoring system by adopting multi-source sensor fusion technology. By synchronously deploying vibration sensors, thermal imaging sensors, current and voltage monitoring units, flow and pressure sensors and environmental parameter monitoring equipment, a comprehensive health status perception system is established, integrating NB-IoT, 5G and wired networks to achieve high-frequency real-time acquisition and cloud-edge collaborative processing of multi-channel data, solving the problems of traditional energy consumption monitoring and health monitoring data dispersion, delayed response and untimely cleaning; through cleaning technologies such as sliding window, wavelet denoising, and dynamic anomaly rejection, the real-time, integrity and high-precision input of the data are guaranteed, effectively improving the overall monitoring response speed and fault diagnosis capabilities of the system.

[0016] 3. The present invention constructs a digital twin virtual-reality synchronization platform to comprehensively map the operating status of electromechanical equipment and the carbon emission change process, and dynamically calculates the carbon emission paths of single devices and system levels in real time. Based on the equipment health index, energy efficiency parameters and carbon emission prediction results, the system can automatically prioritize the scheduling of high-efficiency and low-emission equipment, and dynamically adjust the load distribution strategy to ensure that the equipment operating efficiency and carbon emissions are in the optimal state. Compared with the existing technology that only focuses on carbon emission management in a single operating stage, the present invention supports full life cycle carbon emission monitoring and optimization control, achieving a reduction of more than 9% in carbon emissions of electromechanical systems, and strongly supporting the low-carbon, smart and green operation and maintenance development needs of important infrastructure under the "dual carbon" goals. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of the artificial intelligence-based electromechanical system health monitoring and optimization method of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] like Figure 1 As shown, an embodiment of the present invention provides an artificial intelligence-based electromechanical system health monitoring and optimization method. The specific steps of the artificial intelligence-based electromechanical system health monitoring and optimization method are as follows: S1: Establishing a multi-source sensor network and determining equipment monitoring targets: Based on the finite element analysis model, transient simulation model, and carbon emission calculation model of the building and electromechanical systems, a digital twin infrastructure for the electromechanical system is established. Through model simulation and operating condition analysis, key monitoring targets and representative parts prone to failure are selected, and monitoring points are determined. Targeted at different equipment and environmental parameters, NB-IoT wireless vibration sensors (sampling frequency not less than 500Hz), thermal imaging sensors (resolution not less than 160×120 pixels, refresh rate ≥9Hz), current and voltage sensors (data refresh cycle ≤1s), flow and pressure sensors (response time ≤0.5s), and environmental sensors are deployed. The number of measurement points is configured to be no less than 3 sets of sensor units per 100㎡ of building area to ensure comprehensive data coverage; S2: Real-time data high-frequency acquisition and multi-channel wireless transmission: The collected multi-source monitoring data is asynchronously transmitted in real time via NB-IoT (latency ≤ 2s), 5G (latency ≤ 0.5s), and the RS485 bus (data transmission rate ≥ 115.2kbps). Based on a transmission architecture that combines wireless and wired networks, it collects multiple parameters such as vibration, temperature, current, voltage, flow, pressure, humidity, and illumination during the operation of the electromechanical system at high frequency and in real time. The multi-channel data is then synchronously transmitted to the cloud or edge computing platform, ensuring data continuity, stability, and high timeliness. S3: Health monitoring raw data cleaning and outlier preprocessing: For multi-dimensional data collected in real time, we use data processing techniques such as sliding window analysis, time-domain averaging, and wavelet denoising to complete outlier removal, data completion, and standardization. This ensures the integrity, accuracy, and temporal continuity of health monitoring data and eliminates interference from environmental noise and load fluctuations on subsequent analysis. S4: Electromechanical System Operation Status Feature Extraction and Multi-Scale Trend Analysis: A three-layer convolutional neural network (convolution kernel size 3×3, stride 1, and output channels of 32, 64, and 128, respectively) is used to perform deep feature extraction on four parameters: vibration, thermal, electrical, and fluid. Peak value, RMS value, kurtosis, and spectral peak displacement are extracted from time-domain vibration signals. Total harmonic distortion (THD) and current surge characteristics are extracted from electrical parameters. The maximum temperature difference in high-temperature zones is extracted from thermal imaging. Flow rate fluctuation is extracted from fluid parameters. Equipment operation trends are calculated using time-series data streams to identify potential failure modes. S5: Historical Condition Modeling and Predictive Maintenance Knowledge Base Construction: Based on collected historical equipment condition data, a knowledge base for electromechanical equipment condition monitoring is established. Data-driven support vector machine (SVM) and long short-term memory (LSTM) prediction models are constructed. Input features include: root mean square vibration, harmonic distortion, thermal imaging hot spot temperature, and flow rate fluctuation. The training set must contain ≥ 100,000 data sets, with a prediction error target of less than 5%. The remaining service life and health index of the equipment are output. A knowledge base for equipment failure cases and an operation and maintenance response strategy library are also constructed simultaneously. S6: Real-time status prediction and graded dynamic warning mechanism: Utilizes an LSTM model to predict the equipment's operating status over the next 24 hours, focusing on typical fault signs such as sudden vibration spikes, current distortion, and abnormal temperature rise in thermal imaging. Three levels of dynamic warning thresholds are set: mild anomaly (prediction deviation 5%-10%), moderate anomaly (prediction deviation 10%-15%), and severe anomaly (prediction deviation ≥15%). The warning generation time lag is controlled to ≤1 second, and the system supports alarm linkage and on-site inspection dispatching. S7: Digital Twin Virtual-Real Synchronization and Full Lifecycle Carbon Emissions Mapping: Through the bidirectional integration of finite element simulation models and real-time monitoring data, a digital twin platform for electromechanical systems is established, ensuring virtual-real data synchronization delays of ≤2 seconds and virtual-real residuals of ≤5%. Based on the real-time load, energy efficiency coefficient, and operating time of the equipment, the carbon emissions of each device are dynamically calculated, and a cumulative carbon emissions curve for the full lifecycle of the electromechanical system is generated. The equipment health index and carbon emission forecast results are simultaneously updated; S8: Intelligent optimization scheduling and priority operation strategy implementation: Based on the equipment health index, energy efficiency parameters and carbon emission indicators of the digital twin platform, the equipment start-up and shutdown sequence, load distribution and operation priority are dynamically adjusted. Equipment with good health and higher energy efficiency is prioritized to achieve optimal overall system operation efficiency and minimize carbon emissions, avoiding continuous fatigue of high-load equipment. S9: Cloud-based remote operation and maintenance monitoring and closed-loop feedback optimization: Through the web, app, or local control terminal, equipment health status, early warning information, carbon emissions, and scheduling optimization results are displayed in real time. Synchronize cloud-based operation and maintenance decisions with on-site control systems to form an intelligent operation and maintenance closed-loop management. Combined with new data from subsequent equipment operations, continuous iteration and optimization of prediction models and scheduling strategies enhances the health management capabilities of electromechanical systems throughout their lifecycle. The present invention also continuously collects real-time equipment operation data through cloud-based remote monitoring and intelligent closed-loop feedback mechanisms, supports dynamic iteration and self-optimization of prediction models and scheduling strategies, further enhances the intelligence and adaptability of the system, and has been verified through multi-scenario applications; The system can adapt to complex environments and the operation requirements of multiple types of equipment, and has good engineering feasibility and scalability. It can not only significantly reduce the operation and maintenance costs of electromechanical equipment throughout their life cycle, but also improve the digitalization and intelligence level of equipment management. It provides efficient, low-carbon, and safe intelligent operation and maintenance solutions for important infrastructure such as smart buildings, rail transit, airport hubs, and green parks. It has good promotion prospects and social and economic value.

[0020] The establishment of a multi-source sensor network and the determination of equipment monitoring objects in S1 refer to building a digital twin infrastructure for electromechanical systems based on finite element analysis, transient simulation, and carbon emission calculation models of buildings and electromechanical systems. Through in-depth simulation and operating condition analysis of the models, key monitoring objects and fault-prone parts are accurately screened, and monitoring points are scientifically determined. This provides a reliable basis for the real-time perception and precise control of the subsequent operating status of the electromechanical systems, laying a solid foundation for digital twin applications. Based on differences in equipment characteristics and environmental parameters, multiple types of sensors, such as NB-IoT wireless vibration, thermal imaging, current and voltage, are rationally deployed, and performance parameter requirements for each sensor are clearly defined. Furthermore, the principle of deploying no fewer than three sensor units per 100 square meters of building area is adhered to. In combination with multi-source sensor measurement point optimization logic, the number of measurement points is dynamically planned taking into account factors such as building area to comprehensively ensure the integrity and effectiveness of monitoring data coverage. The high-frequency real-time data acquisition and multi-channel wireless transmission in S2 refers to the use of an asynchronous multi-channel transmission mode that combines NB-IoT, 5G, and RS485 buses to build a data transmission architecture that complements wireless and wired networks. NB-IoT ensures wireless transmission efficiency with a delay of no more than 2s, and 5G ensures low latency of ≤0.5s. The RS485 bus ensures stable wired transmission with a data transmission rate of ≥115.2kbps. The three transmission modes work together to lay a solid and efficient channel foundation for electromechanical system data transmission. Based on this transmission architecture, multiple parameters such as vibration and temperature in the operation of the electromechanical system are collected in real time at high frequency. Using multi-channel data synchronization logic, various types of collected data are quickly and stably transmitted synchronously to the cloud or edge computing platform through different transmission channels. This logic fully considers the characteristics of each transmission channel to ensure high frequency of data collection and continuity, stability, and high timeliness of transmission, providing reliable data support for system analysis and decision-making. Among them, the health monitoring raw data cleaning and outlier preprocessing in S3 refers to the data processing technology of sliding window analysis, time domain averaging, and wavelet denoising for the multi-dimensional data of electromechanical systems collected in real time. The sliding window analysis dynamically captures data features, the time domain averaging is used to integrate data information, and the interference components in the data are filtered out with the help of wavelet denoising. In this process, according to the abnormal data elimination logic, outliers are accurately identified and eliminated, and data completion and standardization are performed at the same time to comprehensively optimize the raw data. Through a series of data processing operations, the integrity, accuracy, and temporal continuity of health monitoring data are effectively ensured. The application of abnormal data elimination logic avoids the impact of outliers on the overall data. Data completion and standardization processing improves data content and unifies the format. These processes eliminate the interference of factors such as environmental noise and load fluctuations on subsequent data analysis, providing a reliable data foundation for data-based electromechanical system status assessment and fault diagnosis. Among them, the electromechanical system operating status feature extraction and multi-scale trend analysis in S4 refers to the use of a three-layer convolutional neural network with a convolution kernel of 3×3, a step size of 1, and output channels of 32, 64, and 128 respectively, to conduct in-depth feature mining on four types of parameters: vibration, heat, electricity, and fluid. For time-domain vibration signals, peak values, root mean square (RMS) and other features are extracted according to specific calculation logic; electrical parameters focus on total harmonic distortion rate and current impact; thermal imaging focuses on the maximum temperature difference in high-temperature areas; and fluid parameters extract flow fluctuation rate to achieve multi-source data feature integration. The extracted multi-parameter features are converted into time-series data streams. Based on the data streams, equipment operation trends are analyzed. Through continuous monitoring and analysis of various parameter features, combined with key indicator changes in the root mean square of the vibration signal, potential equipment failure modes can be accurately identified and the equipment operating status can be predicted in advance, providing strong support for electromechanical system maintenance and fault warning. The historical state modeling and predictive maintenance knowledge base construction in S5 refers to building an electromechanical equipment state monitoring knowledge base based on collected historical equipment state data. Using support vector machines and long short-term memory networks (LSTMs), with vibration root mean square and harmonic distortion rate features as input, a data-driven prediction model is constructed. Deep training is performed using no less than 100,000 training sets of data. The state prediction logic optimization model is optimized using LSTM, strictly controlling the prediction error within 5%, and accurately outputting the remaining service life and health index of the equipment. While completing the construction of the prediction model, we also simultaneously build an equipment fault case knowledge base and an operation and maintenance response strategy base. The fault case knowledge base summarizes historical fault information and provides a reference for problem diagnosis; the operation and maintenance response strategy base formulates targeted treatment plans, achieving full-process knowledge collaboration from fault prediction, diagnosis to disposal, and improving the intelligent operation and maintenance system of electromechanical equipment. The real-time status prediction and hierarchical dynamic warning mechanism in S6 uses the LSTM model to predict the operating status of the equipment for the next 24 hours, focusing on capturing typical fault signs such as sudden vibration increases, current distortion, and abnormal temperature rise in thermal imaging. The LSTM state prediction logic processes time series data to accurately deduce equipment operating trends, providing a reliable prediction basis for fault warning. A three-level dynamic warning threshold system is established: mild anomaly (prediction deviation 5%-10%), moderate anomaly (10%-15%), and severe anomaly (≥15%). The warning generation lag time is controlled within 1 second. This mechanism, combined with dynamic threshold logic, automatically triggers the corresponding warning level based on the prediction deviation. The system also supports alarm linkage and on-site inspection dispatch functions, forming a closed-loop management from prediction to disposal. The S7 digital twin virtual-real synchronization and full-lifecycle carbon emissions mapping refers to the construction of a mechatronic system digital twin platform through the bidirectional integration of finite element simulation models and real-time monitoring data. Relying on virtual-real residual control logic, the quality of data synchronization is strictly controlled to ensure that the virtual-real data synchronization delay is ≤2 seconds and the virtual-real residual is ≤5%, thus achieving accurate mapping and dynamic linkage between the physical system and the virtual model. Based on the equipment's real-time load, energy efficiency coefficient, and operating hours, the platform dynamically calculates the carbon emissions of individual devices and cumulatively generates a carbon emissions curve for the entire life cycle of the electromechanical system. Simultaneously, the platform updates the equipment health index and carbon emissions forecast results in real time, enabling coordinated monitoring and management of equipment operating status and environmental indicators. The implementation of intelligent optimization scheduling and priority operation strategies in S8 refers to formulating dynamic scheduling plans based on the equipment health index, energy efficiency parameters, and carbon emission indicators of the digital twin platform. Based on the health-first scheduling priority logic, the equipment start-up and shutdown sequence, load distribution, and operation priority are determined, giving priority to equipment with good health and high energy efficiency, laying the foundation for efficient system operation. Through this dynamic scheduling, the overall system operating efficiency is optimized and carbon emissions are minimized. Reasonable load distribution avoids continuous fatigue of high-load equipment and extends equipment life. At the same time, while ensuring stable system operation, efficiency and environmental protection are balanced to achieve comprehensive benefits. The cloud-based remote operation and maintenance monitoring and closed-loop feedback optimization in S9 refers to the real-time presentation of equipment health status, early warning information, carbon emission data, and scheduling optimization results through the web, app, and local control terminals. It synchronizes cloud-based operation and maintenance decisions with on-site control systems, follows the time control logic of cloud-based alarm responses, and forms intelligent operation and maintenance closed-loop management to ensure efficient operation and maintenance. By combining new data from equipment operation, we continuously iterate and optimize prediction models and scheduling strategies. By continuously improving model accuracy and strategy adaptability, we gradually enhance the health management capabilities of electromechanical systems throughout their lifecycle, achieving efficient management of the entire process from real-time monitoring and intelligent decision-making to continuous optimization. It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0021] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence-based electromechanical system health monitoring and optimization method, characterized by: This AI-based electromechanical system health monitoring and optimization method builds a digital twin architecture based on simulation models, rationally deploys multi-source sensors, and comprehensively covers key equipment. The system collects operational data at high frequency and transmits it synchronously to the cloud through multiple channels for real-time monitoring. The collected data is cleaned and preprocessed to remove anomalies. The cleaned data is input into a neural network to extract features and identify operational trends and failure modes. A predictive model is established based on historical data, building a maintenance knowledge base. Equipment status is predicted in real time, and early warnings are generated dynamically. Through virtual-real synchronization, health and carbon emission status are mapped in real time, and closed-loop monitoring in the cloud is continuously iterated. The specific steps of the optimization method are as follows: S1: Build a digital twin architecture, optimize monitoring objects and sensor layout, and establish a data collection foundation; S2: Based on the sensor network, collect multi-parameter data in real time and transmit it synchronously to the cloud processing platform; S3: Clean and process the real-time collected data to remove abnormal data; S4: The cleaned data is fed into a neural network to extract operational features and identify trends in equipment status changes. S5: The extracted status features are used for historical modeling to build a predictive model and equipment operation and maintenance knowledge base. S6: The prediction model outputs future equipment status, generates early warning information in real time, and supports dynamic risk monitoring; S7: Early warning data is synchronized to the digital twin platform to map the equipment status and carbon emission changes in real time; S8: Dynamically optimize equipment scheduling based on the twin platform output; S9: Optimization results are synchronized with the cloud platform, remotely monitored, and continuously updated in a closed loop with the model and scheduling strategy.

2. The method for monitoring and optimizing the health of an electromechanical system based on artificial intelligence according to claim 1, characterized in that: The specific steps of establishing a multi-source sensor network and determining the equipment monitoring object in S1 are as follows: Step 1: Digital Twin Architecture Construction and Monitoring Planning: Relying on finite element analysis, transient simulation, and carbon emission calculation models of the building and electromechanical systems, we build a digital twin infrastructure for the electromechanical systems. Through in-depth simulation and operating condition analysis of the model, we accurately select key monitoring objects and fault-prone areas, and scientifically determine monitoring points. Step 2: Multi-source sensor deployment and measurement point optimization: Based on device characteristics and environmental parameter differences, rationally deploy multiple types of NB-IoT wireless vibration, thermal imaging, current and voltage sensors, and clearly define the performance parameter requirements for each sensor. At the same time, adhere to the principle of deploying no fewer than three sensor units per 100 square meters of building area. Combined with the multi-source sensor measurement point optimization logic, the number of measurement points is dynamically planned, taking into account the building area. The optimization expression formula of multi-source sensor measurement points is: ； Where, Minimum number of sensors to be deployed, Monitoring area, Ceiling function.

3. The method for monitoring and optimizing the health of an electromechanical system based on artificial intelligence according to claim 1, characterized in that: The specific steps of high-frequency real-time data acquisition and multi-channel wireless transmission in S2 are as follows: Step 1: Build a multi-channel hybrid transmission architecture: Adopt an asynchronous multi-channel transmission mode that combines NB-IoT, 5G, and the RS485 bus to build a complementary wireless and wired data transmission architecture. NB-IoT ensures wireless transmission efficiency with a latency of no more than 2s, and 5G ensures low latency of ≤0.5s. The RS485 bus ensures stable wired transmission with a data transmission rate of ≥115.2kbps. Step 2: Multi-parameter high-frequency acquisition and data synchronization: Based on the above transmission architecture, high-frequency real-time acquisition of vibration and temperature parameters during electromechanical system operation is performed. Using multi-channel data synchronization logic, the collected data of various types is synchronously transmitted to the cloud or edge computing platform through different transmission channels. The expression formula for multi-channel data synchronization is: ； Where, time Synchronous data collection at all times, No. Data after delayed synchronization of channel sensors, No. Channel transmission delay, The total number of sensor channels. Multi-channel synchronous data needs to be delayed compensated before the timestamp can be unified.

4. The method for monitoring and optimizing the health of an electromechanical system based on artificial intelligence according to claim 1, characterized in that: The specific steps for cleaning the raw health monitoring data and preprocessing abnormal values ​​in S3 are as follows: Step 1: Application of multidimensional data processing technology: For the multidimensional data of electromechanical systems collected in real time, a combination of sliding window analysis, time domain averaging, and wavelet denoising is used. Sliding window analysis dynamically captures data features, time domain averaging is used to integrate data information, and wavelet denoising is used to filter out interference components in the data. In this process, outliers are accurately identified and eliminated based on abnormal data elimination logic, and data completion and standardization are performed simultaneously. Step 2: Data quality assurance and interference elimination: Through data processing operations, effectively ensure the integrity, accuracy and temporal continuity of health monitoring data; The expression formula for abnormal data removal is: ； Where, Standard deviation score (Z-score), Collect data, The mean value in the sliding window, Standard deviation within the sliding window, abnormal data judgment: When |Z|>3, the data is judged to be abnormal and needs to be eliminated.

5. The method for monitoring and optimizing the health of an electromechanical system based on artificial intelligence according to claim 1, characterized in that: The specific steps of the electromechanical system operating state feature extraction and multi-scale trend analysis in S4 are as follows: Step 1: Multi-parameter deep feature extraction: Using a three-layer convolutional neural network with a 3×3 kernel, a stride of 1, and output channels of 32, 64, and 128, we perform deep feature mining on four parameters: vibration, thermal, electrical, and fluid. For time-domain vibration signals, we extract peak and root mean square (RMS) features using specific calculation logic. Electrical parameters focus on total harmonic distortion and current impact; thermal imaging focuses on the maximum temperature difference in high-temperature areas; fluid parameters extract flow fluctuation rate; Step 2: Equipment Failure Pattern Identification: The extracted multi-parameter features are converted into a time-series data stream. Based on the data stream, equipment operation trends are analyzed. Through continuous monitoring and analysis of various parameter features, combined with key indicator changes in the root mean square of the vibration signal, potential equipment failure modes can be accurately identified. The expression formula of the root mean square of the vibration signal is: ； Where, RMS: root mean square value of vibration signal, No. The vibration acceleration of each sampling point, The total number of sampling points.

6. The method for monitoring and optimizing the health of an electromechanical system based on artificial intelligence according to claim 1, characterized in that: The specific steps for constructing the historical state modeling and predictive maintenance knowledge base in S5 are as follows: Step 1: Prediction Model Construction and Training: Based on the collected historical equipment status data, a knowledge base for electromechanical equipment condition monitoring is established. Using support vector machines and long-short-term memory networks, with vibration root mean square and harmonic distortion rate features as input, a data-driven prediction model is constructed. Deep training is performed using no less than 100,000 training data sets. The model is optimized using LSTM state prediction logic to accurately output the equipment's remaining service life and health index. Step 2: Collaborative construction of the knowledge base system: While completing the construction of the prediction model, simultaneously build the equipment failure case knowledge base and operation and maintenance response strategy library. The failure case knowledge base summarizes historical failure information.

7. The method for monitoring and optimizing the health of an electromechanical system based on artificial intelligence according to claim 1, characterized in that: The specific steps of the real-time status prediction and hierarchical dynamic early warning mechanism in S6 are as follows: Step 1: LSTM model trend prediction and fault monitoring: The LSTM model is used to predict the equipment's operating status over the next 24 hours, focusing on typical fault symptoms such as sudden increases in vibration, current distortion, and abnormal temperature rise in thermal imaging. The LSTM state prediction logic processes time series data to accurately predict equipment operating trends. Step 2: Three-level dynamic warning mechanism and system functions: Set up a three-level dynamic warning threshold system: mild anomaly, moderate anomaly, and severe anomaly, and control the warning generation lag time to within 1 second. This mechanism combines dynamic threshold logic to automatically trigger the corresponding warning level based on the prediction deviation. At the same time, the system supports alarm linkage and on-site inspection dispatch functions; The expression formula of the three-level dynamic warning threshold is: ； Where, Percentage of prediction error, Predicted state value, Current actual monitoring status value; Warning level: Mild abnormality: ; Moderate abnormality: ; Severe abnormalities: .

8. The method for monitoring and optimizing the health of an electromechanical system based on artificial intelligence according to claim 1, characterized in that: The specific steps of digital twin virtual-real synchronization and full life cycle carbon emission mapping in S7 are as follows: Step 1: Build a digital twin platform for the electromechanical system: Through the bidirectional integration of finite element simulation models and real-time monitoring data, a digital twin platform for the electromechanical system is constructed. Relying on the virtual-real residual control logic, the quality of data synchronization is strictly controlled to ensure that the virtual-real data synchronization delay is ≤ 2 seconds and the virtual-real residual is ≤ 5%; Step 2: Dynamic management of carbon emissions and health index: Based on the real-time load, energy efficiency coefficient, and operating hours of the equipment, the carbon emissions of each device are dynamically calculated, and a cumulative carbon emissions curve for the entire life cycle of the electromechanical system is generated. At the same time, the platform updates the equipment health index and carbon emission forecast results in real time. The digital twin virtual-real residual control expression is: ； Where, The percentage of virtual and real residuals, Digital twin virtual state data, Real-time actual monitoring data.

9. The method for monitoring and optimizing the health of an electromechanical system based on artificial intelligence according to claim 1, characterized in that: The specific steps for implementing the intelligent optimization scheduling and priority operation strategy in S8 are as follows: Step 1: Dynamic scheduling strategy formulation: Based on the equipment health index, energy efficiency parameters, and carbon emission indicators of the digital twin platform, a dynamic scheduling plan is formulated. Based on the health-first scheduling priority logic, the equipment start-up and shutdown sequence, load distribution, and operation priority are determined; Step 2: Achieve system operation optimization goals: Through the above dynamic scheduling, achieve optimal overall system operation efficiency and minimize carbon emissions; The health-first scheduling priority expression formula is: ； Where, Equipment operation priority score, Equipment health index (0-1), The current energy efficiency coefficient of the equipment (0-1), The device's current carbon emissions per unit time.

10. The method for monitoring and optimizing the health of an electromechanical system based on artificial intelligence according to claim 1, characterized in that: The specific steps of cloud-based remote operation and maintenance monitoring and closed-loop feedback optimization in S9 are as follows: Step 1: Multi-terminal information display and closed-loop operation and maintenance: Through the web, app, and local control terminals, equipment health status, early warning information, carbon emission data, and scheduling optimization results are presented in real time. Cloud-based operation and maintenance decisions are synchronized with the on-site control system, and the time control logic of cloud-based alarm responses is followed. Step 2: Model iteration and management capability improvement: Integrate new equipment operation data to continuously iterate and optimize prediction models and scheduling strategies, gradually improving the health management capabilities of the electromechanical system throughout its life cycle; The cloud alarm response delay expression formula is: ； Where, Total alarm response delay, Data upload transmission delay, Cloud processing computing latency, The cloud-to-client alarm push delay ensures Ttotal ≤ 3 seconds, meeting the cloud-side real-time alarm response requirements.

Images