A multi-source sensing data-based adaptive intelligent early warning system for electromechanical equipment

Abstract

The application belongs to the technical field of electromechanical equipment operation and maintenance, and discloses an electromechanical equipment adaptive intelligent early warning system based on multi-source sensing data, which is composed of a multi-source sensing and perception module, an edge data acquisition and preprocessing module, a data cleaning and multi-dimensional feature extraction module, an equipment state dynamic modeling module, an intelligent fault prediction and trend analysis module, an adaptive early warning threshold generation and dynamic adjustment module, an intelligent decision and remote collaboration module, an intelligent low-carbon operation and maintenance control module, and a digital twin system integration and full-cycle mapping module; the multi-source sensing and perception module is used to deploy various sensors to collect multi-dimensional data of equipment, the edge data acquisition and preprocessing module is used to clean and calibrate the data, the data cleaning and multi-dimensional feature extraction module is used for deep processing, so as to ensure the data integrity and accuracy; the data is quickly transmitted between the modules, so that the monitoring system can present the equipment operation state in real time and accurately.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of electromechanical equipment operation and maintenance, and particularly relates to an adaptive intelligent early warning system for electromechanical equipment based on multi-source sensing data. BACKGROUND

[0002] In modern industrial production, the stable operation of electromechanical equipment is crucial to production efficiency and economic benefits. However, its maintenance and monitoring face many challenges. Currently, the maintenance of electromechanical equipment mainly relies on two modes of post-maintenance and regular maintenance. Post-maintenance is passive maintenance, which is carried out only after the equipment fails and stops. The maintenance timeliness is seriously lagging behind, which not only causes the interruption of the production process, but also may cause a series of chain reactions such as order delay and customer loss due to too long downtime. Regular maintenance is active maintenance, but it uses a fixed periodical maintenance method, which is a one-size-fits-all approach. Regardless of the actual working conditions of the equipment, the same maintenance operation is performed. This not only causes waste of manpower, material resources and financial resources, but also makes it difficult to detect potential failures in advance and effectively prevent equipment accidents due to the lack of accurate judgment of individual differences and real-time state of the equipment.

[0003] In terms of monitoring technology, although some systems introduce single-sensor monitoring methods such as vibration and electrical parameters, the data collected by a single sensor can only reflect the running information of the equipment in one aspect, and it is difficult to fully present the real running state of the equipment, so that the abnormal condition of the equipment cannot be discovered in time and accurately. More importantly, the existing monitoring systems generally ignore the complex nonlinear coupling characteristics between multi-source data, which cannot effectively separate fault signals from non-fault signals, resulting in a high false alarm rate and serious lag in early warning, which cannot meet the stringent requirements of efficient and safe operation of electromechanical equipment. SUMMARY

[0004] The purpose of the present application is to provide an adaptive intelligent early warning system for electromechanical equipment based on multi-source sensing data to solve the problems raised in the background.

[0005] In order to achieve the above-mentioned purpose, the present application provides the following technical solution: an adaptive intelligent early warning system for electromechanical equipment based on multi-source sensing data, which is composed of a multi-source sensing perception module, an edge data acquisition and preprocessing module, a data cleaning and multi-dimensional feature extraction module, an equipment state dynamic modeling module, an intelligent fault prediction and trend analysis module, an adaptive early warning threshold generation and dynamic adjustment module, an intelligent decision-making and remote collaboration module, an intelligent low-carbon operation and management control module, and a digital twin system integration and full-cycle mapping module.

[0006] The multi-source sensing perception module: with the help of wired or wireless communication, a variety of sensors are used to collect multi-source data of electromechanical equipment, including mechanical vibration, electrical parameters, etc., to fully reflect the running state of the equipment and provide raw data for the edge data acquisition and preprocessing module.

[0007] Edge data acquisition and preprocessing module: After receiving the raw data, the preliminary screening, format unification, time synchronization and abnormal signal filtering are completed at the local edge end, the cloud pressure is reduced, the standardized data stream is output, and the data cleaning and multi-dimensional feature extraction module is further processed;

[0008] Data cleaning and multi-dimensional feature extraction module: deep cleaning of preprocessed data, removing interference using multiple methods, extracting time domain, frequency domain and other key feature parameters, forming a multi-dimensional fault feature set, laying a foundation for the equipment state dynamic modeling module;

[0009] Equipment state dynamic modeling module: According to the characteristic parameters, combined with finite element simulation and other technologies to establish a dynamic model, track the equipment state trend, calculate the health index, and provide the intelligent fault prediction and trend analysis module with the basis for fault diagnosis;

[0010] Intelligent fault prediction and trend analysis module: using CNN-LSTM model to integrate health index and historical data, predicting fault trend, outputting health score and fault probability, driving the adaptive warning threshold generation and dynamic adjustment module to adjust the threshold;

[0011] Adaptive warning threshold generation and dynamic adjustment module: based on health score, prediction trend and other data, using Bayesian dynamic regression and other models to generate and update the threshold, and passing the dynamic threshold to the intelligent decision and remote collaboration module to adjust the warning strategy;

[0012] Intelligent decision and remote collaboration module: according to the dynamic threshold to determine the equipment state, push the warning information, support remote interaction and manual intervention, pass the adjusted operation strategy to the intelligent low-carbon operation management and control module, and realize intelligent management and control;

[0013] Intelligent low-carbon operation management and control module: according to the operation strategy to optimize equipment scheduling, reduce energy consumption and prolong life, real-time feedback operation result and equipment state to the digital twin system integration and whole cycle mapping module, assist whole life cycle management;

[0014] Digital twin system integration and whole cycle mapping module: gather multi-module information to build digital twin, realize visualization and analysis function, feedback virtual-real mapping data to the intelligent fault prediction and trend analysis module, form an optimized closed loop.

[0015] Preferably, the multi-source sensing module comprises:

[0016] (1) Data acquisition method and type: Use wired and wireless communication technology to deploy multiple sensors such as vibration, voltage, and thermal imaging to sense the operating state of mechanical and electrical equipment in real time. Collect multi-source data from mechanical vibration, electrical parameters, and thermodynamic state. For example, vibration sensors capture equipment vibration characteristics to determine mechanical component status, and thermal imagers monitor temperature distribution to prevent overheating failures, providing rich data for equipment state analysis;

[0017] (2) Multi-source data acquisition model: Construct a multi-source data acquisition model to process multi-source heterogeneous sensor data using a weighted fusion strategy. This model integrates different types of data, takes full advantage of each sensor, and improves data accuracy and reliability;

[0018] Multi-source data acquisition model formula:

[0019] ;

[0020] In the formula: is the total amount of sensing data of the ith device (multi-source fusion result); is the data collected by the jth sensor of the ith device at time t; is the weight coefficient of the jth sensor of the ith device, set according to the sensor type; is the total number of sensors for each device;

[0021] (3) Data transmission and effect: The collected and fused raw and multi-modal sensor data, which fully reflect the operating state of the equipment, will be transmitted to the edge data acquisition and preprocessing module. These data serve as the basis for subsequent data processing and play a key role in the data processing flow of the entire mechanical and electrical equipment adaptive intelligent early warning system.

[0022] Preferably, the edge data acquisition and preprocessing module includes:

[0023] (1) Data reception and preliminary screening: The edge data acquisition and preprocessing module receives multi-source heterogeneous data uploaded by the multi-source sensing and perception module. These data contain information about the operation of the equipment, but there are errors or invalid records. The edge data acquisition and preprocessing module starts the preliminary screening program, quickly identifies and removes obvious error data according to the preset rules, lays the foundation for subsequent accurate processing, and ensures the reliability of the starting point of data processing;

[0024] (2) Time synchronization preprocessing and accurate calibration: To solve the problem of time synchronization of multi-source sensor data, the edge data acquisition and preprocessing module uses a time synchronization preprocessing formula for accurate calibration, achieving accurate alignment of data in the time dimension and ensuring the accuracy and effectiveness of data analysis;

[0025] Time synchronization preprocessing formula:

[0026] ;

[0027] wherein: time-synchronized sensor data; is the time delay correction value of the jth sensor for the ith device; collected data; is the time delay correction value of the jth sensor;

[0028] (3) Format unification, anomaly filtering and data transmission: after time synchronization, the edge data acquisition and preprocessing module unifies the data formats of different types of sensors, so that the data conforms to the standard specification. At the same time, an anomaly detection algorithm based on statistics and signal characteristics is used to scan the data in real time and eliminate sudden interference and noise. After processing, the standardized data stream is output to the data cleaning and multi-dimensional feature extraction module, which not only ensures data quality but also significantly reduces cloud data processing load.

[0029] Preferably, the data cleaning and multi-dimensional feature extraction module comprises:

[0030] (1) Deep data cleaning operation: this module receives the data after edge preprocessing and starts the deep cleaning process. A denoising algorithm is used to preliminarily remove random noise in the data, and then wavelet denoising filtering is used to analyze the signal in detail at different scales and accurately filter out interference signals according to a specific wavelet denoising filtering principle. Combined with the time domain average method, the multiple collected data are comprehensively processed to effectively enhance the effective signal, eliminate various non-working condition interference and abnormal points, and significantly improve the data quality.

[0031] Wavelet denoising filtering formula:

[0032] ;

[0033] wherein: is the filtered signal; is the wavelet coefficient of the approximate signal; is the wavelet coefficient of the detail signal; is the basis function of scale A; is the basis function of scale j; j is the scale parameter of wavelet decomposition, representing different scale levels of signal analysis, and the value range is [A, J], where A is the starting scale and J is the maximum decomposition scale;

[0034] (2) Multi-dimensional feature parameter extraction: After completing the data deep cleaning, this module starts to extract key feature parameters. Basic statistical features such as mean and variance are extracted from the time domain to characterize the basic characteristics of the signal; the frequency information of the equipment vibration and electrical signal is mined by analyzing the spectral distribution in the frequency domain; through time-frequency domain analysis, the time and frequency dimension information is fused to present the dynamic change of signal frequency with time; for thermal imaging data, features such as temperature extreme value and temperature gradient are extracted to comprehensively evaluate the thermal state of the equipment;

[0035] (3) Fault feature set output and transmission: After data cleaning and multi-dimensional feature extraction, this module integrates to generate a multi-dimensional fault feature set. These feature sets contain key information of the equipment operating state and can accurately reflect the potential fault characteristics of the equipment. The multi-dimensional fault feature set is passed down to the equipment state dynamic modeling module, providing indispensable key data support for subsequent dynamic equipment state modeling, and helping to realize accurate modeling and analysis of equipment state.

[0036] Preferably, the equipment state dynamic modeling module comprises:

[0037] (1) Multi-technology fusion to build dynamic model: Based on the key feature parameters output by the data cleaning and multi-dimensional feature extraction module, this module comprehensively uses finite element simulation, mechanical and electrical system physical modeling, and digital twin real-time synchronization technology. Finite element simulation starts from the details of the equipment structure, discretizes the analysis of mechanical and thermal performance changes under different working conditions; mechanical and electrical system physical modeling constructs mathematical models according to the principle characteristics of the equipment, accurately describes the running process; digital twin real-time synchronization ensures that the virtual model is consistent with the actual equipment state, and together builds a dynamic equipment operating state model;

[0038] (2) Deep tracking of equipment state based on model: The dynamic model established has strong equipment state tracking capability. By continuously analyzing the changes of various parameters in the model, the state trend in the equipment running process can be captured in real time. Combined with the equipment health index model, the equipment state is comprehensively evaluated from multiple dimensions, and the running condition of the equipment is converted into an intuitive and measurable index, and potential performance degradation or abnormal trends are discovered in a timely manner, providing a forward-looking basis for equipment maintenance;

[0039] Equipment health index model formula:

[0040] ;

[0041] In the formula: is the equipment health index (Health Index); is the weight of the th state parameter; is the th state parameter (such as vibration RMS, current, etc.); Total number of state parameters;

[0042] (3) Health index output and subsequent connection: Based on the real-time tracking of equipment state change trend, this module calculates the real-time equipment health index according to the equipment health index model. These indexes directly reflect the current health level of the equipment, and are passed down to the intelligent fault prediction and trend analysis module. As the basis for fault diagnosis, it provides an important reference for the intelligent fault prediction module, helping it to more accurately predict potential equipment failures and promote the efficient operation of the entire early warning system.

[0043] Preferably, the intelligent fault prediction and trend analysis module comprises:

[0044] (1) Deep learning model fusion technology: This module innovatively uses a CNN-LSTM multi-level deep learning model to deeply integrate the real-time equipment health index output by the equipment state model and historical fault data. CNN can automatically mine local features and spatial features in the data due to its strong feature extraction capability, accurately identifying fault patterns hidden in equipment operation data; LSTM has unique advantages in processing time series data, which can effectively learn the long-term change trend of equipment state in the time dimension, and the two complement each other to lay a solid technical foundation for fault prediction;

[0045] (2) Equipment fault trend prediction process: Based on the fusion of data and models, this module deeply predicts the future potential fault trend of the equipment. Through the feature extraction of real-time health indicators and historical data by CNN, combined with the time series analysis of equipment operation trend by LSTM, subtle signals such as equipment performance degradation and parameter abnormal change are captured from massive data, and then the future fault development trend of the equipment is deduced, and the health score and fault probability reflecting the health status of the equipment are dynamically output, providing forward-looking prediction for equipment operation and maintenance;

[0046] (3) Prediction result driving threshold optimization: The equipment fault trend data and fault probability output by this module will be passed down to the adaptive early warning threshold generation and dynamic adjustment module. These prediction results serve as the key basis for dynamically adjusting the early warning threshold, driving the threshold to be optimized in real time according to the actual operation conditions and performance change trend of the equipment. Through this linkage mechanism, the early warning system can closely match the actual operation of the equipment, improve the timeliness and accuracy of the early warning, and realize dynamic and precise control of the operation risk of mechanical and electrical equipment.

[0047] Preferably, the adaptive early warning threshold generation and dynamic adjustment module comprises:

[0048] (1) Multi-source data integration and analysis: This module collects the equipment health score and prediction trend data output by the intelligent fault prediction and trend analysis module, and integrates real-time working condition parameters and environmental change information. These multi-dimensional data comprehensively reflect the equipment operating state, providing a solid data foundation for threshold generation and adjustment, which is the key prerequisite for precise early warning. Through in-depth analysis of the data, subtle changes in the equipment state are captured, providing direction for subsequent threshold adjustment;

[0049] (2) Dynamic model driven threshold generation: Bayesian dynamic regression and exponential smoothing model are used to generate early warning and alarm thresholds based on integrated multi-source data. Bayesian dynamic regression model combines prior knowledge and real-time data to achieve dynamic optimization of thresholds through probabilistic reasoning; the exponential smoothing model uses weighted average of historical data to quickly respond to data trend changes. Based on the principle of dynamic threshold adjustment formula, the two models enable the threshold to closely follow the equipment state fluctuations, always maintaining adaptability to the actual operation of the equipment, improving the accuracy and timeliness of early warning;

[0050] Dynamic threshold adjustment formula:

[0051] ;

[0052] In the formula: is the updated threshold; is the threshold at the previous time is the dynamic adjustment coefficient; is the current predicted failure probability; is the preset baseline probability;

[0053] (3) Threshold transmission and strategy optimization: The generated dynamic threshold is transmitted to the intelligent decision and remote collaboration module in a timely manner, becoming an important basis for equipment state determination. Based on the dynamic threshold, the intelligent decision and remote collaboration module adjusts the early warning strategy in real time, and when the equipment operating parameters reach the threshold, the corresponding early warning measures are triggered quickly. This linkage mechanism ensures that the early warning system can accurately reflect the operating state of the equipment under different working conditions and environments, effectively avoiding false positives and false negatives, and achieving precise control of the operation risk of mechanical and electrical equipment.

[0054] Preferably, the intelligent decision and remote collaboration module comprises:

[0055] (1) Dynamic threshold driven intelligent decision: This module uses the dynamic threshold transmitted by the adaptive early warning threshold generation and dynamic adjustment module as the benchmark to accurately determine the real-time operating state of the mechanical and electrical equipment. Once the equipment operating parameters reach the early warning or alarm threshold, the system triggers the intelligent push mechanism and sends early warning or alarm information of the corresponding level to the relevant personnel, ensuring that equipment abnormalities can be detected in a timely manner and time is gained for subsequent disposal;

[0056] (2) Remote interaction and strategy adjustment: This module supports remote WEB and APP interaction, giving users the ability to remotely control the device. Users can not only view the device's running status in real time and receive warning information, but also manually intervene in the device's running strategy as needed, such as flexibly adjusting operating parameters and switching work modes. At the same time, the system can also automatically optimize the running strategy according to the preset rules, and the adjusted strategy is passed down to the intelligent low-carbon operation and maintenance control module, comprehensively improving the safety and reliability of device operation.

[0057] Preferably, the intelligent low-carbon operation and maintenance control module comprises:

[0058] (1) Strategy-based intelligent operation and maintenance scheduling: This module starts the intelligent operation and maintenance scheduling mechanism based on the adjustment strategy output by the intelligent decision and remote collaboration module. High-efficiency devices are preferentially enabled, and the load is scientifically and reasonably distributed according to device performance and work requirements, while devices with high energy consumption and low efficiency are automatically shut down. Through optimization and adjustment of device running combinations and working parameters, unnecessary energy consumption is reduced, the overall energy consumption of the system is effectively reduced on the premise of ensuring normal operation of the device, and the service life of the device is prolonged;

[0059] (2) Carbon emission optimization and data feedback: During the intelligent operation and maintenance scheduling process, the system carbon emission optimization model is used to calculate and accurately control the carbon emissions during the operation and maintenance phase. This model takes into account multiple factors such as device running energy consumption and working time, dynamically monitors carbon emissions, and optimizes operation and maintenance strategies accordingly. At the same time, the low-carbon operation results and device running status are fed back to the digital twin system integration and full-cycle mapping module in real time, providing detailed data for device full-life cycle management and helping to achieve the goal of low-carbon and intelligent operation and maintenance of mechanical and electrical equipment.

[0060] System carbon emission optimization model formula:

[0061] ;

[0062] In the formula: is the total carbon emissions of the current system; is the carbon emission factor of the i-th device; is the power load of the i-th device; is the time step; is the total number of devices participating in carbon emission calculation in the system; Preferably, the digital twin system integration and full-cycle mapping module comprises:

[0063] Preferably, the digital twin system integration and full-cycle mapping module comprises:

[0064] ​(1) Multi-source data fusion to construct a digital twin: This module real-time gathers the equipment operation data collected by the multi-source sensing and sensing module, the fault prediction results of the intelligent fault prediction and trend analysis module, and the operation state information of the intelligent low-carbon operation and control module. Based on this, a highly realistic digital twin of the equipment is constructed, which completely maps the full life cycle state of the equipment. Through three-dimensional visualization, the details of the equipment operation are presented, the evolution of the fault is analyzed by means of historical data backtracking, intelligent optimization analysis is carried out to propose operation and maintenance schemes, and the overall control of the equipment state is realized;

[0065] (2) Error correction driven closed-loop optimization: During the operation of the digital twin, the virtual-real synchronization error correction formula is used to monitor and dynamically correct the deviation between the actual state of the equipment and the virtual model in real time, so as to ensure that the virtual model accurately reflects the real condition of the equipment. The corrected virtual-real mapping data of the equipment is fed back to the fault prediction and threshold adjustment module to provide more accurate data support, forming a full life cycle adaptive optimization closed loop, and continuously improving the system warning accuracy and the scientific nature of equipment operation and maintenance;

[0066] Virtual-real synchronization error correction formula:

[0067] ;

[0068] In the formula: is the virtual-real synchronization error; is the real state parameter of the equipment; is the digital twin prediction state parameter.

[0069] The beneficial effects of the present application are as follows:

[0070] 1. The present application deploys multiple sensors through the multi-source sensing and sensing module to collect multi-dimensional data of the equipment, cleans and calibrates the data through the edge data collection and preprocessing module, and deeply processes the data through the data cleaning and multi-dimensional feature extraction module, so as to ensure the data integrity and accuracy; the digital twin system integrates the real-time data gathering and mapping module to construct a virtual model, realizes the comprehensive perception of the equipment state, and realizes the rapid transmission of data between modules, the edge computing cooperation from data collection to preliminary processing, greatly shortens the response time, makes the monitoring system able to present the equipment operation state in real time and accurately, provides reliable basis for subsequent decision-making, and effectively improves the real-time performance and data integrity of the monitoring system.

[0071] 2、The application constructs a device operation model by combining finite element simulation and digital twin technology through the device state dynamic modeling module, the intelligent fault prediction and trend analysis module adopts a CNN-LSTM model, and fuses device health indicators and historical fault data; the CNN identifies fault modes, the LSTM learns state trends, and the combination of the two accurately excavates data characteristics to predict potential faults in advance; the physical model assists in verifying the prediction results, avoids misjudgments caused by single data or algorithms, significantly reduces the false positive rate and the false negative rate, realizes accurate early warning of device faults, and ensures the safe and stable operation of mechanical and electrical equipment.

[0072] 3、The adaptive early warning threshold generation and dynamic adjustment module generates an early warning threshold dynamically based on the results of the intelligent fault prediction and trend analysis module, combines device operating conditions and environmental data, and uses Bayesian dynamic regression or exponential smoothing model; this mechanism breaks the limitations of fixed thresholds and adjusts them in real time according to the actual operating state of the device. When the device operating conditions change or the environment changes, the threshold is optimized to ensure that the early warning strategy is accurately adapted; the intelligent decision-making and remote collaboration module determines the device state in a timely manner based on the dynamic threshold to achieve accurate early warning, and comprehensively enhances the intelligence and adaptability of the system, improving the operation and maintenance efficiency and reliability of mechanical and electrical equipment. BRIEF DESCRIPTION OF DRAWINGS

[0073] Figure 1 The application is a process diagram of a mechanical and electrical equipment adaptive intelligent early warning system based on multi-source sensing data. DETAILED DESCRIPTION

[0074] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the application.

[0075] As shown in Figure 1 The application provides a mechanical and electrical equipment adaptive intelligent early warning system based on multi-source sensing data, which is composed of a multi-source sensing perception module, an edge data acquisition and preprocessing module, a data cleaning and multi-dimensional feature extraction module, a device state dynamic modeling module, an intelligent fault prediction and trend analysis module, an adaptive early warning threshold generation and dynamic adjustment module, an intelligent decision-making and remote collaboration module, an intelligent low-carbon operation and management control module, and a digital twin system integration and full-cycle mapping module;

[0076] Multi-source sensing perception module: with the help of wired or wireless communication, a variety of sensors are used to collect multi-source data of mechanical and electrical equipment, including mechanical vibration, electrical parameters, etc., to comprehensively reflect the operating state of the equipment and provide raw data for the edge data acquisition and preprocessing module;

[0077] Edge data acquisition and preprocessing module: After receiving raw data, preliminary screening, format unification, time synchronization and abnormal signal filtering are completed at the local edge end, reducing the pressure on the cloud, and outputting standardized data stream for data cleaning and multi-dimensional feature extraction module for further processing;

[0078] Data cleaning and multi-dimensional feature extraction module: deep cleaning of preprocessed data, removing interference using multiple methods, extracting key feature parameters such as time domain and frequency domain, forming a multi-dimensional fault feature set, laying a foundation for the equipment state dynamic modeling module;

[0079] Equipment state dynamic modeling module: based on feature parameters, combined with finite element simulation technology to establish a dynamic model, track equipment state trends, calculate health indicators, and provide fault diagnosis basis for intelligent fault prediction and trend analysis module;

[0080] Intelligent fault prediction and trend analysis module: using CNN-LSTM model to integrate health indicators and historical data to predict fault trends, output health scores and fault probabilities, and drive the adaptive warning threshold generation and dynamic adjustment module to adjust the threshold;

[0081] Adaptive warning threshold generation and dynamic adjustment module: based on health scores, prediction trends and other data, use Bayesian dynamic regression model to generate and update thresholds, pass the dynamic threshold to the intelligent decision and remote collaboration module to adjust the warning strategy;

[0082] Intelligent decision and remote collaboration module: determine the equipment state according to the dynamic threshold, push the warning information, support remote interaction and manual intervention, pass the adjusted operation strategy to the intelligent low-carbon operation and control module to realize intelligent control;

[0083] Intelligent low-carbon operation and control module: optimizes equipment scheduling according to operation strategy, reduces energy consumption and prolongs life, and feeds back operation results and equipment state to the digital twin system integration and full-cycle mapping module to assist full-life cycle management;

[0084] Digital twin system integration and full-cycle mapping module: integrates multi-module information to build a digital twin, realizes visualization and analysis functions, and feeds back virtual and real mapping data to the intelligent fault prediction and trend analysis module to form an optimized closed loop.

[0085] Based on multi-source sensing data, an adaptive intelligent early warning system for electromechanical equipment is constructed to realize efficient operation and maintenance and accurate early warning. At the data perception level, the multi-source sensing perception module deploys multiple sensors to collect multi-dimensional data of the equipment. After processing by the edge data collection and preprocessing module, the data cleaning and multi-dimensional feature extraction module, the data is ensured to be complete and accurate. The digital twin system integration and full-cycle mapping module constructs a virtual model to achieve comprehensive perception of the equipment state and rapid data transmission to improve the real-time monitoring of the equipment. For fault prediction, the equipment state dynamic modeling module integrates finite element simulation and digital twin technology, and the intelligent fault prediction and trend analysis module uses a CNN-LSTM model to combine equipment health indicators and historical data to accurately mine fault characteristics and reduce the false alarm rate. In the early warning mechanism, the adaptive early warning threshold generation and dynamic adjustment module dynamically generates thresholds based on multi-source data to break the limitations of fixed thresholds. The intelligent decision-making and remote collaboration module accurately determines the equipment state to enhance the intelligence and adaptability of the system and ensure reliable operation of the equipment.

[0086] The multi-source sensing perception module comprehensively uses wired and wireless communication technologies to deploy various types of sensors such as vibration, voltage, and thermal imaging to perceive the running state of electromechanical equipment in real time and from multiple dimensions such as mechanical vibration, electrical parameters, and thermodynamic state. For example, vibration sensors can capture equipment vibration characteristics to determine the condition of mechanical parts, and thermal imagers can monitor temperature distribution to prevent overheating failures.

[0087] A weighted fusion strategy is used to process multi-source heterogeneous data to fully leverage the advantages of each sensor and effectively improve the accuracy and reliability of the data. The original and multi-modal sensor data collected and fused present the complete running state of the equipment and are transmitted to the edge data collection and preprocessing module to provide indispensable input for subsequent data preprocessing, feature extraction, and state modeling, supporting the efficient operation of the entire early warning system.

[0088] The edge data collection and preprocessing module mainly undertakes the preprocessing of multi-source heterogeneous data. The module first receives data uploaded by the multi-source sensing perception module. Although these data cover various aspects of equipment operation, they may contain errors or invalid records. The module quickly eliminates invalid data based on preset rules through an initial screening program to ensure the reliability of data processing.

[0089] To address the problem of time asynchronization of multi-source sensor data, a time synchronization preprocessing formula is used for accurate calibration to achieve accurate alignment of data in the time dimension and provide effective protection for subsequent analysis.

[0090] The edge data acquisition and preprocessing module unifies the data format, conforms to the standard specification, and uses an abnormality detection algorithm based on statistics and signal characteristics to filter out sudden interference and noise in real time, and outputs the standardized data stream to the data cleaning and multi-dimensional feature extraction module, which greatly reduces the cloud data processing load while ensuring data quality, and lays a solid foundation for the data processing of the entire early warning system.

[0091] The data cleaning and multi-dimensional feature extraction module is mainly responsible for deep processing of the data after edge preprocessing. The module first uses a denoising algorithm to preliminarily remove random noise, and then analyzes the signal from different scales based on the wavelet denoising filter principle to accurately filter out interference signals;

[0092] At the same time, combined with the time domain average method, the effective signal is enhanced and the abnormal data is removed, which significantly improves the data quality; after deep cleaning, the module extracts key feature parameters from multiple dimensions such as time domain, frequency domain, time-frequency domain and thermal imaging data, and fully describes the equipment operating state;

[0093] The multi-dimensional fault feature set obtained by cleaning and extraction is integrated and output, and is transmitted to the equipment state dynamic modeling module to provide core data for building a dynamic equipment state model and laying a solid foundation for accurately judging the equipment state and predicting potential faults.

[0094] The equipment state dynamic modeling module undertakes the task of accurate modeling of the equipment state and output of key indicators. The module takes the key feature parameters output by the data cleaning and multi-dimensional feature extraction module as the cornerstone, and integrates finite element simulation, mechanical and electrical system physical modeling, and digital twin real-time synchronization technology. Finite element simulation analyzes the performance changes under equipment working conditions from structural details, physical modeling accurately describes the running process based on principles, and digital twin ensures real-time synchronization between virtual model and equipment state, which together build a dynamic equipment operating state model;

[0095] Based on this model, the module continuously tracks the equipment state in depth, comprehensively evaluates the equipment operating condition from multiple dimensions with the help of the equipment health index model, and captures the performance degradation or abnormal trend in real time. The real-time equipment health index calculated finally is transmitted downward to the intelligent fault prediction and trend analysis module to provide core basis for fault diagnosis and help the early warning system accurately predict potential equipment faults.

[0096] The intelligent fault prediction and trend analysis module makes forward-looking judgments and early warning optimization for equipment faults. The module innovatively uses a CNN-LSTM multi-level deep learning model to deeply integrate the real-time health index output by the equipment state model and historical fault data, and plays the ability of CNN to automatically extract local and spatial features and accurately identify fault patterns, and combines the advantages of LSTM in processing time series data and learning long-term trends of equipment state;

[0097] Deeply excavate the subtle abnormal signal in the equipment operation data, deduce the potential failure trend in the future, and dynamically output the health score and failure probability. The prediction results are transmitted to the adaptive warning threshold generation and dynamic adjustment module to drive the threshold to be optimized in real time according to the equipment working condition and performance trend, so as to realize the precise control of the mechanical and electrical equipment operation risk and improve the timeliness and accuracy of the early warning system.

[0098] Among them, the adaptive warning threshold generation and dynamic adjustment module first collects the equipment health score and prediction trend data output by the intelligent fault prediction and trend analysis module, and integrates real-time working condition parameters and environmental change information of the equipment. Through in-depth analysis of multi-dimensional data, subtle changes in the state of the equipment are accurately captured;

[0099] Adopting Bayesian dynamic regression and exponential smoothing model, combining with the principle of dynamic threshold adjustment formula, the warning and alarm thresholds are generated according to the integrated data, so that the threshold can closely fit the dynamic changes of the equipment operation state. Finally, the generated dynamic threshold is transmitted to the intelligent decision and remote collaboration module as the basis for equipment state judgment, driving the real-time adjustment of the warning strategy to ensure that the early warning system can effectively avoid false positives and false negatives under different working conditions and environments.

[0100] Among them, the intelligent decision and remote collaboration module accurately judges the real-time operation state of the equipment based on the dynamic threshold transmitted by the adaptive warning threshold generation and dynamic adjustment module. When the equipment parameters reach the threshold, the intelligent push mechanism is triggered to send early warning or alarm information in time.

[0101] Supporting remote WEB and APP end interaction, users can not only master the equipment state in real time and receive warnings, but also manually intervene in the operation strategy. The system can also automatically optimize the strategy according to the preset rules. The adjusted strategy is transmitted to the intelligent low-carbon operation and control module to comprehensively improve the safety and reliability of equipment operation.

[0102] Among them, the intelligent low-carbon operation and control module starts intelligent operation and maintenance scheduling according to the adjustment strategy output by the intelligent decision and remote collaboration module, preferentially enables high-efficiency equipment, reasonably allocates load, shuts down low-efficiency equipment, optimizes equipment operation combination and parameters, and effectively reduces energy consumption and prolongs equipment life while ensuring normal operation of the equipment.

[0103] With the help of the system carbon emission optimization model, the carbon emissions during operation and maintenance are calculated and accurately controlled in real time by comprehensively considering factors such as equipment energy consumption and working time, and the operation and maintenance strategy is dynamically optimized accordingly. At the same time, the low-carbon operation results and equipment operation state are fed back to the digital twin system integration and whole cycle mapping module to provide data support for the whole life cycle management of the equipment.

[0104] The digital twin system integrates and maps the real-time multi-source sensing data of the device operation data of the sensing module, the prediction results of the intelligent fault prediction and trend analysis module, and the operation state information of the intelligent low-carbon operation and management module, constructs a high-fidelity digital twin body, and realizes three-dimensional visualization of the device life cycle state, historical backtracking and intelligent optimization.

[0105] During operation, the deviation between the actual state of the device and the virtual model is monitored and corrected in real time by means of the virtual-real synchronization error correction formula, so as to ensure accurate mapping and to feed back the corrected virtual-real mapping data to the intelligent fault prediction and trend analysis module, forming a closed loop optimization and continuously improving the system warning accuracy and operation scientificity.

[0106] It should be noted that, in this article, relationship terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device.

[0107] Although the embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A multi-source sensor data based adaptive intelligent early warning system for electromechanical equipment, characterized in that: The system is composed of a multi-source sensing perception module, an edge data acquisition and preprocessing module, a data cleaning and multi-dimensional feature extraction module, a device state dynamic modeling module, an intelligent fault prediction and trend analysis module, an adaptive early warning threshold generation and dynamic adjustment module, an intelligent decision and remote collaboration module, an intelligent low-carbon operation management and control module, and a digital twin system integration and full-cycle mapping module. The multi-source sensing perception module uses sensors to collect multi-source data of mechanical and electrical equipment, including mechanical vibration, electrical parameters, and obtains raw data. The edge data acquisition and preprocessing module performs initial screening, format unification, time synchronization, and abnormal signal filtering on the received multi-source data, and outputs the preprocessed data to the data cleaning and multi-dimensional feature extraction module. The data cleaning and multi-dimensional feature extraction module deeply cleans the preprocessed data, removes interference, extracts time-domain and frequency-domain key feature parameters, and forms a multi-dimensional fault feature set. The device state dynamic modeling module constructs a device dynamic operation model based on the multi-dimensional fault feature set, calculates the device health index in real time, and transmits it to the intelligent fault prediction and trend analysis module. The intelligent fault prediction and trend analysis module fuses the device health index and historical data, predicts the device fault trend through a deep learning model, generates a device health score and a fault probability, and transmits them to the adaptive early warning threshold generation and dynamic adjustment module in real time. The adaptive early warning threshold generation and dynamic adjustment module generates a dynamic threshold based on the health score and predicted trend data using a Bayesian dynamic regression model. The intelligent decision and remote collaboration module determines the device operating state based on the dynamic threshold, generates an operation adjustment strategy, and pushes it to the intelligent low-carbon operation management and control module in real time, while pushing early warning information to the remote management end to support remote interaction and manual intervention. The intelligent low-carbon operation management and control module optimizes device scheduling based on the operation strategy, and feeds back operation results and device status to the digital twin system integration and full-cycle mapping module in real time. The digital twin system integration and full-cycle mapping module constructs a device digital twin, monitors the virtual-real synchronization error in real time, and corrects the virtual-real mapping data after correction and returns it to the intelligent fault prediction and trend analysis module, forming a closed-loop optimization control chain.

2. The adaptive intelligent early warning system for electromechanical equipment based on multi-source sensing data according to claim 1, characterized in that: The multi-source sensing perception module includes: (1) Data acquisition method and type: use wired and wireless communication technology to deploy vibration, voltage, and thermal imaging sensors to sense the operating state of mechanical and electrical equipment in real time, from mechanical vibration, electrical parameters to thermodynamic state, and collect multi-source data in all directions; (2) Multi-source data acquisition model: construct a multi-source data acquisition model, use a weighted fusion strategy to process multi-source heterogeneous sensor data, and integrate different types of data through the model; Multi-source data acquisition model formula: ； In the formula: is the total amount of sensor data of the jth device; is the total amount of sensor data of the jth device; is the data collected by the jth sensor of the jth device at time t; is the data collected by the jth sensor of the jth device at time t; is the weight coefficient of the jth sensor of the jth device; is the weight coefficient of the jth sensor of the jth device; is the total number of sensors of each device; (3) Data transmission and effect: after collection and fusion, the original and multi-modal sensor data are transmitted to the edge data acquisition and preprocessing module. 3.The adaptive intelligent early warning system for electromechanical equipment based on multi-source sensing data according to claim 1, characterized in that: The edge data acquisition and preprocessing module includes: (1) Data reception and preliminary screening: there are error or invalid records in the received multi-source heterogeneous data, start the preliminary screening program, and quickly identify and eliminate obvious error data according to the preset rules; (2) Time synchronization preprocessing and accurate calibration: In order to solve the problem of time asynchronization of multi-source sensor data, time synchronization preprocessing formula is used for accurate calibration. Time synchronization preprocessing formula: ； In the formula: synchronized time data; is the time delay correction value of the jth sensor for the ith device; collected data; is the time delay correction value of the jth sensor. (3) Format unification and abnormality filtering: After time synchronization, the edge data acquisition and preprocessing module unifies the data format of different types of sensors, and uses an abnormality detection algorithm based on statistics and signal characteristics to scan the data in real time and eliminate sudden interference and noise.

4. The adaptive intelligent early warning system for electromechanical equipment based on multi-source sensing data according to claim 1, characterized in that: The data cleaning and multi-dimensional feature extraction module includes: (1) Deep data cleaning operation: After receiving the data processed by the edge data acquisition and preprocessing module, a denoising algorithm is used to preliminarily remove random noise in the data, and then wavelet denoising filtering is used to analyze the signal in detail at different scales and accurately filter out interference signals. Wavelet denoising filtering formula: ； wherein: is the filtered signal; is the wavelet coefficient of the approximation signal; is the wavelet coefficient of the detail signal; is the basis function of scale A; is the basis function of scale j; j is a scale parameter of wavelet decomposition, and its value range is [A, J], wherein A is a starting scale, and J is a maximum decomposition scale. (2) Multi-dimensional feature parameter extraction: After deep data cleaning, the frequency information of equipment vibration and electrical signals is extracted by analyzing the frequency spectrum distribution in the frequency domain; the time and frequency dimension information is fused by analyzing the time-frequency domain; for thermal imaging data, temperature extreme value and temperature gradient features are extracted; (3) Fault feature set output and transmission: After data cleaning and multi-dimensional feature extraction, a multi-dimensional fault feature set is generated and transmitted to the equipment state dynamic modeling module.

5. The adaptive intelligent early warning system for electromechanical equipment based on multi-source sensing data according to claim 1, characterized in that: The equipment state dynamic modeling module includes: (1) Multi-technology fusion to build dynamic model: Based on the key feature parameters output by the data cleaning and multi-dimensional feature extraction module, finite element simulation, mechanical and electrical system physical modeling, and digital twin real-time synchronization technology are comprehensively used to analyze the changes of mechanical and thermal performance under different working conditions through finite element simulation discretization; the mathematical model is constructed according to the principle characteristics of the equipment through mechanical and electrical system physical modeling; digital twin real-time synchronization ensures that the virtual model is consistent with the actual equipment state; (2) Deep tracking of equipment state based on model: By analyzing the parameter changes in the model, combining with the equipment health index model, the running condition of the equipment is converted into an intuitive and measurable index, and potential performance degradation or abnormal trend is discovered in time; Equipment health index model formula: ； In the formula: is the equipment health index; is the weight of the th state parameter; is the th state parameter; is the total number of state parameters; (3) Health index output and subsequent connection: The real-time equipment health index is calculated according to the equipment health index model, and these indexes are transmitted to the intelligent fault prediction and trend analysis module as the basis for fault diagnosis.

6. The adaptive intelligent early warning system for electromechanical equipment based on multi-source sensing data according to claim 1, characterized in that: The intelligent fault prediction and trend analysis module includes: (1) Deep learning model fusion technology: CNN-LSTM deep learning model is used to deeply integrate real-time equipment health index and historical fault data; (2) Equipment fault trend prediction process: The future potential fault trend of the equipment is deeply predicted through CNN feature extraction of real-time health index and historical data, combined with LSTM time series analysis of equipment operation trend, and then the future fault development trend of the equipment is deduced; (3) Prediction result driven threshold optimization: The equipment fault trend data and fault probability are transmitted to the adaptive warning threshold generation and dynamic adjustment module, and the driven threshold is optimized in real time according to the actual operation condition and performance change trend of the equipment.

7. The adaptive intelligent early warning system for electromechanical equipment based on multi-source sensing data according to claim 1, characterized in that: The adaptive warning threshold generation and dynamic adjustment module includes: (1) Multi-source data integration and analysis: Collect the equipment health score and prediction trend data output by the intelligent fault prediction and trend analysis module, and integrate real-time working condition parameters and environmental change information of the equipment. Through in-depth analysis of the data, subtle changes in the state of the equipment are captured; (2) Dynamic model driven threshold generation: Bayesian dynamic regression model realizes dynamic optimization of threshold through probability reasoning; exponential smoothing model uses weighted average of historical data to quickly respond to data trend changes. Based on the principle of dynamic threshold adjustment formula, the threshold can closely follow the fluctuations of the equipment state. Dynamic threshold adjustment formula: ； In the formula, T is the threshold value, T is the updated threshold value, T is the threshold value of the previous moment, T is the dynamic adjustment coefficient, P is the current predicted failure probability, and P is the preset baseline probability. is an updated threshold value; is a threshold value of a previous moment; is a dynamic adjustment coefficient; is a current predicted failure probability; is a preset baseline probability; (3) Threshold transmission and strategy optimization: The generated dynamic threshold is transmitted to the intelligent decision and remote collaboration module. Based on the dynamic threshold, the intelligent decision and remote collaboration module adjusts the early warning strategy in real time. When the equipment operating parameters reach the threshold, the corresponding early warning measures are triggered quickly. 8.The adaptive intelligent early warning system for electromechanical equipment based on multi-source sensing data of claim 1, wherein: The intelligent decision and remote collaboration module includes: (1) Dynamic threshold driven intelligent decision: Based on the received dynamic threshold, the real-time running state of the mechanical and electrical equipment is determined. When the equipment operating parameters reach the warning or alarm threshold, the system triggers the intelligent push mechanism immediately; (2) Remote interaction and strategy adjustment: Users can view the equipment operating state in real time, receive warning information, and manually intervene in the equipment operation strategy. The system automatically optimizes the operation strategy according to the preset rules, and the adjusted strategy is transmitted to the intelligent low-carbon operation and control module.

9. The adaptive intelligent early warning system for electromechanical equipment based on multi-source sensing data according to claim 1, characterized in that: The intelligent low-carbon operation and control module includes: (1) Strategy-based intelligent operation scheduling: According to the received adjustment strategy, high-efficiency equipment is preferentially enabled, and load is allocated according to equipment performance and work demand, while automatically shutting down equipment with high energy consumption and low efficiency; (2) Carbon emission optimization and data feedback: With the system carbon emission optimization model, real-time calculation and precise control of carbon emissions during operation and maintenance are performed, and the operation and maintenance strategy is optimized accordingly. The low-carbon operation results and equipment operating state are fed back to the digital twin system integration and full-cycle mapping module in real time; System carbon emission optimization model formula: ； wherein: is the total amount of carbon emission of the current system; is the carbon emission factor of the ith device; is the power load of the ith device; is the time step; n is the total number of devices in the system that participate in the carbon emission calculation.

10. The adaptive intelligent early warning system for electromechanical equipment based on multi-source sensing data according to claim 1, characterized in that: The digital twin system integration and full-cycle mapping module includes: (1) Multi-source data fusion to build digital twin: Real-time aggregation of equipment operation data, fault prediction results, and operation status information to build a digital twin of the equipment, which fully maps the full life cycle state of the equipment; (2) Error correction driven closed-loop optimization: During the operation of the digital twin, the virtual-real synchronization error correction formula is used to monitor and dynamically correct the deviation between the actual state of the equipment and the virtual model in real time. The corrected virtual-real mapping data of the equipment is fed back to the intelligent fault prediction and trend analysis module; Virtual-real synchronization error correction formula: ； In the formula: is a virtual-real synchronization error; is a real state parameter of the device; is a digital twin predicted state parameter.

Images