Large machine lubrication state real-time monitoring and early warning system based on Internet of Things

Abstract

The invention discloses a large-scale machine lubrication state real-time monitoring and early warning system based on the Internet of Things, relates to the technical field of state real-time monitoring and early warning, and solves the problems that firstly, the large-scale machine lubrication state is difficult to monitor in real time by using a sensor network of the Internet of Things; secondly, after the lubrication state is difficult to intelligently recognize and classify, early warning is carried out through time sequence analysis; thirdly, simulating a flowing path of the lubricant and analyzing an influence coefficient on flowing of the lubricant according to a fault identification model and a digital twinning technology, and carrying out early warning; and finally, the technical problem that the early warning information is difficult to display visually by using a digital twinborn model is solved. The lubrication state is monitored in real time through the sensor, data are transmitted to the cloud through the Internet of Things, the lubrication state is analyzed through the deep learning model after the data are processed through the server, simulation analysis early warning is conducted in combination with the digital twinborn technology, and finally early warning information is visually displayed through the digital twinborn model.

Description

Technical Field

[0001] The present invention belongs to the technical field of real-time status monitoring and early warning, and specifically is a real-time monitoring and early warning system for lubrication status of a large machine based on the Internet of Things. Background Art

[0002] In industrial production, the lubrication status of large-scale mechanical equipment, such as heavy engineering machinery and high-speed electric spindles, is directly related to its operating efficiency, service life and safety. IoT technology realizes real-time perception, dynamic monitoring and intelligent decision-making of the physical world through sensor networks and data transmission. Through IoT technology, the lubrication status parameters of equipment can be monitored in real time; by integrating deep learning algorithms and digital twin technologies, it is possible to achieve comprehensive, accurate and real-time monitoring and early warning of the lubrication status of large equipment.

[0003] The following problems exist in the existing technology: first, it is difficult to use the sensor network of the Internet of Things to monitor the lubrication status of large machines in real time; second, it is difficult to intelligently identify and classify the lubrication status and issue an early warning through time series analysis; then, it is difficult to simulate the flow path of the lubricant and analyze the influence coefficient on the lubricant flow based on the fault identification model combined with digital twin technology to issue an early warning; finally, it is difficult to use the digital twin model to visualize the early warning information. Summary of the invention

[0004] In order to solve the problems existing in the above-mentioned prior art, the first aspect of the present invention provides a real-time monitoring and early warning system for the lubrication status of a large machine based on the Internet of Things, comprising the following modules: Data acquisition module: monitors the lubrication status of large machines in real time through sensors and collects monitoring data in real time; Data transmission module: using the IoT communication protocol, the monitoring data collected by the sensor is transmitted to the cloud server through a wireless network or a wired network; Data processing module: Store, clean and standardize monitoring data on cloud servers; Data analysis and early warning module: Use the deep belief network of the deep learning model to intelligently identify and classify the lubrication status; perform time series analysis on the lubrication status based on the identification and classification results of the current lubrication status, and issue early warnings by calculating and predicting the lubrication status value under the lubrication status; combine digital twin technology to build a digital twin model of the lubrication status of the large machine to simulate the lubrication status of the large machine; establish a fault identification model based on the monitoring data; simulate the flow path of the lubricant based on the fault identification results combined with the digital twin model, and issue early warnings by calculating the influence coefficient of the fault point on the lubricant flow; Visual warning display module: Visually display warning information through the digital twin model.

[0005] Further: the lubrication status of the large machine is monitored in real time by sensors, and monitoring data is collected in real time, including the following steps: Various sensors are installed in various parts of the large machine's lubrication system, including bearings, gears, oil tanks, oil pumps, oil pipes, and lubrication points. Among them, temperature sensors are installed to monitor the temperature of large machine equipment and lubricant liquid; pressure sensors are installed to monitor the pressure of the lubricant; oil quality sensors are installed to monitor the density, conductivity and dielectric constant of the lubricant; vibration sensors are installed to monitor the vibration information of the equipment; noise sensors are installed to monitor the noise signal of the equipment; moisture sensors are installed to monitor the moisture content in the lubricant; the viscosity of the lubricant is monitored in real time and a particle counter is used to monitor the number of particles in the lubricant; laser displacement sensors are introduced to monitor the oil film thickness of the lubricant in real time.‌ Further: using the deep belief network of the deep learning model to intelligently identify and classify the lubrication state, including the following steps: Based on the processed monitoring data, the deep belief network of the deep learning model is used to automatically learn the feature representation of the monitoring data, identify and classify the lubrication state; the categories of lubrication state include: fluid dynamic lubrication, liquid static lubrication, elastic fluid dynamic lubrication, thin film lubrication, boundary lubrication, dry friction and mixed lubrication; The recognition and classification model is constructed by using the deep belief network model; the restricted Boltzmann machine is used as the basic building block of the deep belief network model; each layer of the restricted Boltzmann machine is trained separately through the layer-by-layer greedy training algorithm; when training each layer of the restricted Boltzmann machine, the contrastive divergence algorithm is used to optimize the parameters of the deep belief network model; the trained restricted Boltzmann machines are stacked layer by layer to form a complete deep belief network model; After the layer-by-layer greedy training is completed, the back propagation algorithm is used to globally optimize the entire deep belief network model; a softmax classifier is added to the output layer of the deep belief network model to classify the lubrication state; Collect historical monitoring data to generate a training data set, and annotate the training data set to label the corresponding lubrication status category labels; Use the labeled training data set to input into the deep belief network model for training, including the training of the softmax classifier; The monitoring data collected and processed in real time is input into the trained deep belief network model, and the identification and classification results of the current large machine lubrication status at the time point of the real-time collected large machine monitoring data are output, including: fluid dynamic lubrication, liquid static lubrication, elastic fluid dynamic lubrication, thin film lubrication, boundary lubrication, dry friction and mixed lubrication.

[0006] Further: according to the identification and classification results of the current lubrication state, a time series analysis of the lubrication state is performed, and an early warning is given by calculating and predicting the lubrication state value under the lubrication state, including the following steps: The current lubrication state is identified based on the trained deep belief network model; by continuously monitoring the lubrication system of the large machine, it is identified whether there is a transition from one lubrication state to another; When there is a transition, the monitoring data of the lubrication status of the large machine at different time points are collected, and the collected monitoring data are arranged in chronological order to form a time series data set; A long short-term memory network model is constructed, and the time series data set is input into the long short-term memory network model for training; the lubrication state of the next time step is predicted by inputting the monitoring data of the current lubrication state identified by the deep belief network model into the trained long short-term memory network model; The sliding window method is used to gradually predict the lubrication status of multiple time steps in the future. Starting from the current window of real-time monitoring, the long short-term memory network model is used to predict the lubrication status of the next time step; by sliding the window forward one time step, adding the latest real-time monitoring data, removing the earliest monitoring data, and repeating the prediction process; by continuously sliding the window and using the long short-term memory network model for prediction, the lubrication status of multiple time steps in the future is gradually obtained; For the predicted lubrication state of multiple time steps in the future, the monitoring data under the predicted lubrication state is obtained; the lubrication state value formula under the predicted lubrication state is calculated: , in, Indicates the lubrication status value, represents the weight of the i-th sensor, represents the monitoring value of the i-th sensor at the j-th time step, represents the minimum value of the historical monitoring value of the i-th sensor, represents the maximum value of the historical monitoring value of the i-th sensor; t is the length of the time window; A lubrication status life end threshold is set. When the lubrication status value is lower than the threshold, the life end is reached and the early warning mechanism is triggered to issue an early warning.

[0007] Further: combining digital twin technology, constructing a digital twin model of the lubrication state of the large machine, simulating the lubrication state of the large machine, including the following steps: According to the actual geometric shape and size of the large machine lubrication system, a geometric model is constructed using 3D modeling software, which describes the various components of the lubrication system, including: the shape, position and interconnection relationship of the bearings, gears, oil tanks, oil pumps, oil pipes, and lubrication points; Based on the constructed geometric model, the flow state of lubricant in the large machine lubrication system is simulated by solving the fluid dynamics equation, including oil pressure distribution and flow velocity distribution; the temperature distribution of lubricant in the large machine lubrication system and the heat transfer process are obtained by solving the heat conduction equation; the distribution of oil pressure, oil film thickness and friction parameters in the lubrication system is obtained by solving the mechanical equilibrium equation; By integrating the above geometric models with the results of fluid dynamics, heat conduction and mechanical balance analysis, a digital twin model of the lubrication status of the large machine was constructed; The monitoring data of the large machine lubrication system is collected in real time through IoT sensors, and the digital twin model is associated with the monitoring data collected by the IoT sensors. The actual operating status of the large machine lubrication system in the digital twin model is synchronized in real time through the data interface; The trained deep belief network model is deployed into the digital twin model, the current real-time lubrication status is output through the deep belief network model, and simulated and displayed through the digital twin model.

[0008] Further: establishing a fault identification model based on the monitoring data, including the following steps: The fault recognition model is established by combining the convolutional neural network model in deep learning with the recurrent neural network model; The historical monitoring data set of the lubrication status of the large machine is collected and processed, and the fault types are marked, including leakage and blockage; the annotated historical monitoring data set is input into the fault recognition model for training; The monitoring data collected and processed through real-time monitoring of the Internet of Things is input into the trained fault identification model, which outputs whether a fault is identified under the current lubrication state of the large machine and the type of fault.

[0009] Further: according to the fault identification result combined with the digital twin model, the flow path of the lubricant is simulated, and an early warning is issued by calculating the influence coefficient of the fault point on the lubricant flow, including the following steps: Deploy the trained fault identification model to the digital twin model, and identify the fault points in the lubricant flow path, including leakage points and blockage points, based on the fault identification model combined with the digital twin model. According to the identification results, count the total number K of all fault points in the flow path. The flow path of the lubricant in the large machine equipment is simulated according to the digital twin model and divided into a normal flow path and an abnormal propagation path; the abnormal propagation path is the flow path of the large machine lubrication state output by the fault identification model under the fault; During the simulation, computational fluid dynamics methods are used to collect the flow conditions of the lubricant on the normal flow path and the abnormal propagation path, including: flow velocity, temperature and pressure distribution; The influence coefficient formula of the fault point on the lubricant flow is calculated by: , in, is the influence coefficient, is the pressure under the kth fault point, is the flow rate at the kth fault point, is the temperature at the kth fault point, is the pressure of the normal flow path, is the flow velocity of the normal flow path, is the temperature of the normal flow path, , and are the rates of change of pressure, flow rate and temperature at the kth fault point over time, respectively, and K represents the total number of all fault points on the flow path; when the influence coefficient exceeds the threshold of 0.2, the early warning mechanism is triggered to issue an early warning.

[0010] Further: visually displaying the warning information through the digital twin model, including the following steps: based on the calculated results of the lubrication state value under the predicted lubrication state and the influence coefficient of the fault point on the lubricant flow, when the warning mechanism is triggered, the generated warning information is integrated into the digital twin model, and the warning information includes: warning type, warning location and warning time.

[0011] Compared with the prior art, the present invention has the following beneficial effects: The present invention monitors the lubrication status of large machines in real time through a variety of sensors, and can comprehensively and accurately obtain various monitoring data of the lubrication system of large machine equipment; by utilizing the Internet of Things communication protocol to transmit the monitoring data to the cloud server, the stability, efficiency and security of data transmission are ensured, and the concurrent data transmission of large-scale equipment is supported. It is also compatible with wireless networks and wired networks, reducing the risk of data transmission interruption or loss.

[0012] After processing the monitoring data on a cloud server, the present invention uses a deep belief network model to intelligently identify and classify the lubrication status, and combines time series analysis and prediction technology to obtain the lubrication status value under the predicted lubrication status and provide early warning based on it.

[0013] The present invention uses digital twin technology to construct a digital twin model of the lubrication status of a large machine, visually presenting the equipment lubrication system; combines the fault identification model with the digital twin model to simulate the lubricant flow path; and issues an early warning by analyzing the influence coefficient of the fault point on the flow path on the lubricant flow. Finally, the digital twin model is used to visualize the early warning information, realizing comprehensive monitoring and early warning of the lubrication status of the large machine. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0015] Figure 1 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0016] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] See also Figure 1 The first embodiment of the present invention provides a real-time monitoring and early warning system for lubrication status of a large machine based on the Internet of Things, including the following modules: Data acquisition module: monitors the lubrication status of large machines in real time through sensors and collects monitoring data in real time; Data transmission module: using the IoT communication protocol, the monitoring data collected by the sensor is transmitted to the cloud server through a wireless network or a wired network; Data processing module: Store, clean and standardize monitoring data on cloud servers; Data analysis and early warning module: Use the deep belief network of the deep learning model to intelligently identify and classify the lubrication status; perform time series analysis on the lubrication status based on the identification and classification results of the current lubrication status, and issue early warnings by calculating and predicting the lubrication status value under the lubrication status; combine digital twin technology to build a digital twin model of the lubrication status of the large machine to simulate the lubrication status of the large machine; establish a fault identification model based on the monitoring data; simulate the flow path of the lubricant based on the fault identification results combined with the digital twin model, and issue early warnings by calculating the influence coefficient of the fault point on the lubricant flow; Visual warning display module: Visually display warning information through the digital twin model.

[0018] Specifically, according to the characteristics of the lubrication system of the large machine, select the appropriate sensor type, including but not limited to: temperature sensor, pressure sensor, flow sensor, etc. By installing sensors at the key lubrication parts of the large machine, ensure that the lubrication status can be fully and accurately monitored. The sensor collects lubrication status data in real time, and adopts lightweight IoT communication protocols such as MQTT and CoAP to ensure low power consumption, real-time and reliability of data transmission. According to the on-site environment, select the appropriate network transmission method, including: in areas covered by wired networks, give priority to using wired networks; in areas covered by wireless networks, use wireless networks such as 4G / 5G and Wi-Fi. During the data transmission process, encryption technologies such as TLS / SSL are used to ensure data security and prevent data leakage or tampering. Establish a database on the cloud server to store the collected monitoring data. The database design should meet the needs of efficient data storage, retrieval and analysis. Perform preprocessing operations such as denoising, deduplication, and filling missing values ​​on the received monitoring data to improve data quality, and standardize the data to ensure that the data collected by different sensors are consistent and comparable. Use the deep belief network model to intelligently identify and classify the lubrication status. By training the deep belief network model, it can accurately identify different lubrication states. Perform trend analysis on the time series data of the lubrication state to predict the future lubrication state. According to the prediction results, analyze the lubrication state value under the predicted lubrication state to issue an early warning. Combined with the actual structure of the large machine and the characteristics of the lubrication system, a digital twin model of the lubrication state of the large machine is constructed, which should be able to simulate the lubrication system. A fault identification model is established based on the monitoring data to identify potential fault points. Combined with the digital twin model, the flow path of the lubricant is simulated, and the influence coefficient of the fault point on the flow of the lubricant is calculated. When the influence coefficient exceeds the preset threshold, an early warning is triggered. Finally, the early warning information is visualized through the digital twin model.

[0019] In this embodiment, the lubrication status of the large machine is monitored in real time by a sensor, and monitoring data is collected in real time, including the following steps: Various sensors are installed in various parts of the lubrication system of the large machine, including bearings, gears, oil tanks, oil pumps, oil pipes, and lubrication points. Among them, temperature sensors are installed to monitor the temperature of the large machine equipment and lubricant liquid; pressure sensors are installed to monitor the pressure of the lubricant; oil quality sensors are installed to monitor the density, conductivity and dielectric constant of the lubricant; vibration sensors are installed to monitor the vibration information of the equipment; noise sensors are installed to monitor the noise signal of the equipment; moisture sensors are installed to monitor the moisture content in the lubricant; the viscosity of the lubricant is monitored in real time and a particle counter is used to monitor the number of particles in the lubricant; laser displacement sensors are introduced to monitor the oil film thickness of the lubricant in real time.

[0020] Specifically, determine the installation location and quantity of sensors according to the structure of large equipment and the characteristics of the lubrication system. Use appropriate tools and methods to fix the sensors in the predetermined position to ensure the stability and accuracy of the sensors. Connect the signal line and power line of the sensors to ensure that the sensors can work normally. After the sensors are installed, debug and calibrate them to ensure the measurement accuracy and reliability of the sensors. By comparing with the actual measured values ​​or standard values, adjust the parameters such as the sensitivity and zero point of the sensors.

[0021] Temperature sensor: Install temperature sensors on key parts of large machines such as bearings, gears, and oil tanks to comprehensively monitor the temperature status of the equipment. Install pressure sensors at oil pump outlets, oil pipes, and other locations to monitor the pressure changes of the lubricant in real time. Install oil quality sensors in oil tanks or oil pipes to obtain lubricant quality information in real time. Install vibration sensors on key transmission parts such as bearings and gears to monitor the vibration of the equipment. Install noise sensors on or near the equipment casing to obtain noise information of the equipment in real time. Monitor the moisture content in the lubricant to prevent lubrication performance degradation and equipment corrosion caused by excessive moisture. Install moisture sensors in oil tanks or oil pipes to monitor the moisture content in the lubricant in real time. Monitor the viscosity of the lubricant through a dedicated viscometer or the viscosity monitoring function integrated in the oil quality sensor. Install a particle counter in the oil tank or oil pipe to monitor the number of particles in the lubricant in real time. Install laser displacement sensors near key lubrication points including bearings, gears, etc. to measure the thickness of the oil film.‌ In this embodiment, the deep belief network of the deep learning model is used to intelligently identify and classify the lubrication state, including the following steps: Based on the processed monitoring data, the deep belief network of the deep learning model is used to automatically learn the feature representation of the monitoring data, identify and classify the lubrication state; the categories of lubrication state include: fluid dynamic lubrication, liquid static lubrication, elastic fluid dynamic lubrication, thin film lubrication, boundary lubrication, dry friction and mixed lubrication; The recognition and classification model is constructed by using the deep belief network model; the restricted Boltzmann machine is used as the basic building block of the deep belief network model; each layer of the restricted Boltzmann machine is trained separately through the layer-by-layer greedy training algorithm; when training each layer of the restricted Boltzmann machine, the contrastive divergence algorithm is used to optimize the parameters of the deep belief network model; the trained restricted Boltzmann machines are stacked layer by layer to form a complete deep belief network model; After the layer-by-layer greedy training is completed, the back propagation algorithm is used to globally optimize the entire deep belief network model; a softmax classifier is added to the output layer of the deep belief network model to classify the lubrication state; Collect historical monitoring data to generate a training data set, and annotate the training data set to label the corresponding lubrication status category labels; Use the labeled training data set to input into the deep belief network model for training, including the training of the softmax classifier; The monitoring data collected and processed in real time is input into the trained deep belief network model, and the identification and classification results of the current large machine lubrication status at the time point of the real-time collected large machine monitoring data are output, including: fluid dynamic lubrication, liquid static lubrication, elastic fluid dynamic lubrication, thin film lubrication, boundary lubrication, dry friction and mixed lubrication.

[0022] Specifically, a deep belief network model is used as the recognition and classification model, and a restricted Boltzmann machine is used as the basic building block of the deep belief network model. Through the layer-by-layer greedy training algorithm, the restricted Boltzmann machine of each layer is first trained separately. When training each layer of the restricted Boltzmann machine, the contrast divergence algorithm is used to optimize the parameters of the deep belief network model. The trained restricted Boltzmann machines are stacked layer by layer to form a complete deep belief network model. After the layer-by-layer greedy training is completed, the back propagation algorithm is used to perform global optimization on the entire deep belief network model to further improve the performance of the model. A softmax classifier is added to the output layer of the deep belief network model to classify the lubrication state. The softmax classifier can convert the output of the model into a probability distribution, thereby realizing accurate classification of different lubrication states. Collect historical monitoring data, which should include monitoring data under different lubrication states. The collected training data set is labeled, and the corresponding category labels of the lubrication state are labeled, including: fluid dynamic lubrication, liquid static lubrication, elastic fluid dynamic lubrication, thin film lubrication, boundary lubrication, dry friction and mixed lubrication. Preprocess the collected data, including data cleaning, normalization and other operations to ensure data quality and consistency. Use the labeled training data set to input the deep belief network model for training, including the training of the softmax classifier. During the training process, attention should be paid to suppressing overfitting. Techniques such as data enhancement, Dropout mechanism, L1 / L2 regularization, etc. can be used to improve the generalization ability of the model; grid search, random search or Bayesian optimization methods can be used to find the optimal hyperparameter combination; by adjusting the model parameters and hyperparameters such as learning rate, the performance and accuracy of the model can be optimized. After the training is completed, the model should be evaluated and verified using the validation data set to ensure the accuracy and reliability of the model. At the same time, the performance of the model can be further evaluated by methods such as cross-validation. Through the monitoring data collected and processed in real time, these data should contain the same features and information as the training data set. The processed real-time monitoring data is input into the trained deep belief network model. The model will identify and classify the lubrication state according to the learned feature representation, and output the identification and classification results of the lubrication state, including: fluid dynamic lubrication, liquid static lubrication, elastic fluid dynamic lubrication, thin film lubrication, boundary lubrication, dry friction and mixed lubrication.

[0023] In this embodiment, according to the identification and classification results of the lubrication state, a time series analysis of the lubrication state is performed, and an early warning is given by calculating and predicting the lubrication state value under the lubrication state, including the following steps: The current lubrication state is identified based on the trained deep belief network model; by continuously monitoring the lubrication system of the large machine, it is identified whether there is a transition from one lubrication state to another; When there is a transition, the monitoring data of the lubrication status of the large machine at different time points are collected, and the collected monitoring data are arranged in chronological order to form a time series data set; A long short-term memory network model is constructed, and the time series data set is input into the long short-term memory network model for training; the lubrication state of the next time step is predicted by inputting the monitoring data of the current lubrication state identified by the deep belief network model into the trained long short-term memory network model; The sliding window method is used to gradually predict the lubrication status of multiple time steps in the future. Starting from the current window of real-time monitoring, the long short-term memory network model is used to predict the lubrication status of the next time step; by sliding the window forward one time step, adding the latest real-time monitoring data, removing the earliest monitoring data, and repeating the prediction process; by continuously sliding the window and using the long short-term memory network model for prediction, the lubrication status of multiple time steps in the future is gradually obtained; For the predicted lubrication state of multiple time steps in the future, the monitoring data under the predicted lubrication state is obtained; the lubrication state value formula under the predicted lubrication state is calculated: , in, Indicates the lubrication status value, represents the weight of the i-th sensor, represents the monitoring value of the i-th sensor at the j-th time step, represents the minimum value of the historical monitoring value of the i-th sensor, represents the maximum value of the historical monitoring value of the i-th sensor; t is the length of the time window; A lubrication status life end threshold is set. When the lubrication status value is lower than the threshold, the life end is reached and the early warning mechanism is triggered to issue an early warning.

[0024] Specifically, historical monitoring data of the machine lubrication system is collected, including various monitoring data, timestamps and other information in sensors under different lubrication conditions. The deep belief network model is trained using the collected historical data. The monitoring data of the current machine lubrication system is identified in real time using the trained deep belief network model to determine the current lubrication state. By continuously monitoring the machine lubrication system, it is identified whether there is a transition from one lubrication state to another. When the state transition is detected, the monitoring data of the machine lubrication state at different time points is collected. The collected monitoring data are arranged in chronological order to form a time series data set. This data set will be used to train the long short-term memory network model. A long short-term memory network model is constructed, which can process time series data and capture long-term dependencies in the data. The time series data set is input into the long short-term memory network model for training, so that the model can learn the law of lubrication state changes over time. The monitoring data of the current lubrication state identified by the deep belief network model is input into the trained long short-term memory network model to predict the lubrication state of the next time step. The sliding window method is used to gradually predict the lubrication state of multiple time steps in the future. Starting from the current window of real-time monitoring, the long short-term memory network model is used to predict the lubrication state of the next time step. Then, the window is slid forward by one time step, the latest real-time monitoring data is added, the earliest monitoring data is removed, and the prediction process is repeated. According to the predicted lubrication state of multiple future time steps, the lubrication state value of the monitoring data under the predicted lubrication state is calculated. By setting a lubrication state life end threshold, a large amount of monitoring data of the lubrication state of the large machine at different time points is first collected, including data when it is in normal operation and when lubrication problems occur; then the collected historical data is analyzed to determine the changing trend and characteristics of the lubrication state when the equipment is in normal operation and when a fault occurs; then, according to the analysis results, a suitable threshold setting method is selected, including statistical methods such as mean ± standard deviation, methods based on equipment operation experience such as referring to the maintenance manual provided by the equipment manufacturer, and machine learning methods such as using historical data to train a classification model to distinguish between normal and abnormal states; then the set threshold is applied to the actual monitoring data, the triggering of the early warning mechanism is observed, and the threshold is verified to accurately reflect the life end of the lubrication state. If the early warning is too frequent or untimely, the threshold needs to be adjusted. In this embodiment, the life end threshold of the lubrication state is set to 0.3, and the specific threshold needs to be adjusted and determined according to the above method and actual conditions. When the calculated lubrication state value is lower than the threshold, it is considered that the large machine equipment has reached the end of its life, and the early warning mechanism needs to be triggered for early warning.

[0025] In this embodiment, the digital twin technology is combined to construct a digital twin model of the lubrication state of the large machine, and the lubrication state of the large machine is simulated, including the following steps: According to the actual geometric shape and size of the large machine lubrication system, a geometric model is constructed using 3D modeling software, which describes the various components of the lubrication system, including: the shape, position and interconnection relationship of the bearings, gears, oil tanks, oil pumps, oil pipes, and lubrication points; Based on the constructed geometric model, the flow state of lubricant in the large machine lubrication system is simulated by solving the fluid dynamics equation, including oil pressure distribution and flow velocity distribution; the temperature distribution of lubricant in the large machine lubrication system and the heat transfer process are obtained by solving the heat conduction equation; the distribution of oil pressure, oil film thickness and friction parameters in the lubrication system is obtained by solving the mechanical equilibrium equation; By integrating the above geometric models with the results of fluid dynamics, heat conduction and mechanical balance analysis, a digital twin model of the lubrication status of the large machine was constructed; The monitoring data of the large machine lubrication system is collected in real time through IoT sensors, and the digital twin model is associated with the monitoring data collected by the IoT sensors. The actual operating status of the large machine lubrication system in the digital twin model is synchronized in real time through the data interface; The trained deep belief network model is deployed into the digital twin model, the current real-time lubrication status is output through the deep belief network model, and simulated and displayed through the digital twin model.

[0026] Specifically, the actual geometric shape, size and detailed parameters of each component of the large machine lubrication system, such as bearings, gears, oil tanks, oil pumps, oil pipes, lubrication points, etc., are collected, and a geometric model of the large machine lubrication system is constructed based on the collected data using 3D modeling software such as SolidWorks and AutoCAD. The shape, position and interconnection relationship of each component are described in detail in the geometric model to ensure that the model is consistent with the actual situation. Based on the constructed geometric model, the flow state of the lubricant in the large machine lubrication system is simulated by solving the fluid dynamics equation, and the oil pressure distribution and flow velocity distribution are analyzed to understand the flow of the lubricant in the system. By solving the heat conduction equation, the temperature distribution of the lubricant in the large machine lubrication system and the heat transfer process are obtained, which helps to evaluate the thermal performance of the lubrication system and ensure that the lubricant works within a suitable temperature range. By solving the mechanical equilibrium equation, the distribution of key parameters such as oil pressure, oil film thickness, and friction parameters in the lubrication system is obtained. These parameters are of great significance for evaluating the performance of the lubrication system and predicting potential failures. The geometric model and the results of fluid dynamics, heat conduction and mechanical balance analysis are integrated to construct a digital twin model of the lubrication status of the large machine, which should be able to fully reflect the actual situation of the large machine lubrication system, including the geometric shape, positional relationship and physical characteristics of each component. By designing a data interface, real-time data synchronization between the digital twin model and the IoT sensor is achieved. The IoT sensor is used to collect monitoring data of the large machine lubrication system in real time. The monitoring data of the large machine lubrication system is collected in real time through the IoT sensor and the collected data is synchronized to the digital twin model in real time through the data interface to reflect the actual operating status of the large machine lubrication system. The trained deep confidence network model is used to predict the real-time lubrication status of the large machine lubrication system. The trained deep confidence network model is deployed in the digital twin model, and the current real-time lubrication status is output through the deep confidence network model; the digital twin model is used for simulation and display to intuitively present the lubrication status of the large machine lubrication system.

[0027] In this embodiment, establishing a fault identification model based on monitoring data includes the following steps: The fault recognition model is established by combining the convolutional neural network model in deep learning with the recurrent neural network model; The historical monitoring data set of the lubrication status of the large machine is collected and processed, and the fault types are marked, including leakage and blockage; the annotated historical monitoring data set is input into the fault recognition model for training; The monitoring data collected and processed through real-time monitoring of the Internet of Things is input into the trained fault identification model, which outputs whether a fault is identified under the current lubrication state of the large machine and the type of fault.

[0028] Specifically, collect historical monitoring data of the lubrication status of large machines, which should cover different operating conditions and fault conditions. Ensure the integrity, accuracy and consistency of the data for subsequent analysis and processing. Clean the collected historical monitoring data to remove noise, outliers and missing values; normalize or standardize the data to improve the convergence speed and performance of the model. Label the monitoring data according to the fault type, including but not limited to: leakage and blockage, to provide supervision information for model training. Use the convolutional neural network model in deep learning combined with the recurrent neural network model to establish a fault recognition model. The convolutional neural network model is good at extracting local features from data with strong spatial correlation, while the recurrent neural network model is suitable for processing time series data and capturing time dependencies. A network structure containing multiple convolutional layers and pooling layers is designed in the convolutional neural network model to extract spatial features from lubrication status monitoring data. The recurrent neural network model is connected to the recurrent neural network after the convolutional neural network model to process time series data and capture the pattern of fault development over time. Set the corresponding output node according to the fault type, and use the softmax function for multi-classification. Input the annotated historical monitoring data set into the fault recognition model for training. The cross entropy loss function is used as the optimization target, and the model parameters are updated using optimization algorithms such as gradient descent. The training process and performance of the model are optimized by adjusting hyperparameters such as learning rate and batch size. The lubrication status of the large machine is monitored in real time using the Internet of Things technology, and real-time monitoring data is collected. The real-time data is preprocessed to maintain the data format and feature space consistent with the training data. The preprocessed real-time monitoring data is input into the trained fault identification model. The model outputs whether a fault is identified under the current lubrication status of the large machine, and gives specific fault types, including leakage or blockage.

[0029] In this embodiment, the flow path of the lubricant is simulated according to the fault identification result combined with the digital twin model, and an early warning is issued by calculating the influence coefficient of the fault point on the lubricant flow, including the following steps: Deploy the trained fault identification model to the digital twin model, and identify the fault points in the lubricant flow path, including leakage points and blockage points, based on the fault identification model combined with the digital twin model. According to the identification results, count the total number K of all fault points in the flow path. The flow path of the lubricant in the large machine equipment is simulated according to the digital twin model and divided into a normal flow path and an abnormal propagation path; the abnormal propagation path is the flow path of the large machine lubrication state output by the fault identification model under the fault; During the simulation, computational fluid dynamics methods are used to collect the flow conditions of the lubricant on the normal flow path and the abnormal propagation path, including: flow velocity, temperature and pressure distribution; The influence coefficient formula of the fault point on the lubricant flow is calculated by: , in, is the influence coefficient, is the pressure under the kth fault point, is the flow rate at the kth fault point, is the temperature at the kth fault point, is the pressure of the normal flow path, is the flow velocity of the normal flow path, is the temperature of the normal flow path, , and are the rates of change of pressure, flow rate and temperature at the kth fault point over time, respectively, and K represents the total number of all fault points on the flow path; when the influence coefficient exceeds the threshold of 0.2, the early warning mechanism is triggered to issue an early warning.

[0030] Specifically, the trained fault identification model is seamlessly integrated into the digital twin model, which ensures that the fault identification module can obtain data from the digital twin model in real time and output fault point information including leakage points and blockage points. The integrated fault identification model is used in combination with the real-time data of the digital twin model to accurately identify whether there is a fault point in the lubricant flow path. The fault point information will be used for subsequent lubricant flow simulation and influence coefficient calculation. In the digital twin model, the complete flow path of the lubricant in the large machine equipment is simulated, including: normal flow path and abnormal propagation path caused by fault. According to the output of the fault identification model, the lubricant flow path is divided into a normal flow path and an abnormal propagation path. The abnormal propagation path specifically refers to the flow path of the lubricant under the fault state. During the simulation process, the computational fluid dynamics method is used to accurately collect the temperature, flow velocity and pressure distribution data of the lubricant on the normal flow path and the abnormal propagation path. The collected temperature, flow velocity and pressure distribution data are substituted into the influence coefficient formula to understand the specific impact of all fault points on the lubricant flow. In this embodiment, the threshold of the influence coefficient is set to 0.2, and the threshold of the influence coefficient is dynamically adjusted by combining actual conditions, expert experience, simulation experiments or historical data. When the calculated influence coefficient exceeds this threshold, the early warning mechanism is triggered to issue an early warning.

[0031] In this embodiment, the warning information is visualized through a digital twin model, including the following steps: based on the calculated results of the lubrication state value under the predicted lubrication state and the influence coefficient of the fault point on the lubricant flow, when the warning mechanism is triggered, the generated warning information is integrated into the digital twin model, and the warning information includes: warning type, warning location and warning time.

[0032] Specifically, when the calculated and predicted lubrication state value is lower than the threshold, the end of life is reached and the early warning mechanism is triggered; when the fault identification model detects the fault point in combination with the digital twin model and calculates that the influence coefficient of the fault point on the lubricant flow exceeds the preset threshold, the early warning mechanism is triggered. According to the above early warning mechanism, early warning information is generated, including early warning types including: lubrication state reaches the end of life, leakage point warning, blockage warning, etc., the early warning location is the specific location of the fault point and the early warning time is the precise time point when the early warning is triggered. The generated early warning information is seamlessly integrated into the digital twin model to ensure that the early warning information can be synchronized and accurately displayed with the real-time data of the digital twin model.

[0033] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A real-time monitoring and early warning system for lubrication status of large machines based on the Internet of Things, characterized in that: Includes the following modules: Data acquisition module: monitors the lubrication status of large machines in real time through sensors and collects monitoring data in real time; Data transmission module: using the IoT communication protocol, the monitoring data collected by the sensor is transmitted to the cloud server through a wireless network or a wired network; Data processing module: Store, clean and standardize monitoring data on cloud servers; Data analysis and early warning module: Use the deep belief network of the deep learning model to intelligently identify and classify the lubrication status; perform time series analysis on the lubrication status based on the identification and classification results of the current lubrication status, and issue an early warning by calculating and predicting the lubrication status value under the lubrication status; Combined with digital twin technology, a digital twin model of the lubrication status of the large machine is constructed to simulate the lubrication status of the large machine; a fault identification model is established based on the monitoring data; based on the fault identification results combined with the digital twin model, the flow path of the lubricant is simulated, and an early warning is issued by calculating the influence coefficient of the fault point on the lubricant flow; Visual warning display module: Visually display warning information through the digital twin model.

2. According to the Internet of Things-based real-time monitoring and early warning system for lubrication status of large machines according to claim 1, it is characterized in that: The lubrication status of the large machine is monitored in real time by sensors, and monitoring data is collected in real time, including the following steps: Various sensors are installed in various parts of the lubrication system of the large machine, including bearings, gears, oil tanks, oil pumps, oil pipes, and lubrication points. Among them, temperature sensors are installed to monitor the temperature of the large machine equipment and lubricant liquid; pressure sensors are installed to monitor the pressure of the lubricant; oil quality sensors are installed to monitor the density, conductivity and dielectric constant of the lubricant; vibration sensors are installed to monitor the vibration information of the equipment; noise sensors are installed to monitor the noise signal of the equipment; moisture sensors are installed to monitor the moisture content in the lubricant; the viscosity of the lubricant is monitored in real time and a particle counter is used to monitor the number of particles in the lubricant; laser displacement sensors are introduced to monitor the oil film thickness of the lubricant in real time.

3. According to the Internet of Things-based real-time monitoring and early warning system for lubrication status of large machines according to claim 1, it is characterized in that: The deep belief network of the deep learning model is used to intelligently identify and classify the lubrication status, including the following steps: Based on the processed monitoring data, the deep belief network of the deep learning model is used to automatically learn the feature representation of the monitoring data, identify and classify the lubrication state; the categories of lubrication state include: fluid dynamic lubrication, liquid static lubrication, elastic fluid dynamic lubrication, thin film lubrication, boundary lubrication, dry friction and mixed lubrication; The recognition and classification model is constructed by using the deep belief network model; the restricted Boltzmann machine is used as the basic building block of the deep belief network model; each layer of the restricted Boltzmann machine is trained separately through the layer-by-layer greedy training algorithm; when training each layer of the restricted Boltzmann machine, the contrastive divergence algorithm is used to optimize the parameters of the deep belief network model; the trained restricted Boltzmann machines are stacked layer by layer to form a complete deep belief network model; After the layer-by-layer greedy training is completed, the back propagation algorithm is used to globally optimize the entire deep belief network model; a softmax classifier is added to the output layer of the deep belief network model to classify the lubrication state; Collect historical monitoring data to generate a training data set, and annotate the training data set to label the corresponding lubrication status category labels; Use the labeled training data set to input into the deep belief network model for training, including the training of the softmax classifier; The monitoring data collected and processed in real time is input into the trained deep belief network model, and the identification and classification results of the current large machine lubrication status at the time point of the real-time collected large machine monitoring data are output, including: fluid dynamic lubrication, liquid static lubrication, elastic fluid dynamic lubrication, thin film lubrication, boundary lubrication, dry friction and mixed lubrication.

4. According to the Internet of Things-based real-time monitoring and early warning system for lubrication status of large machines according to claim 3, it is characterized in that: According to the identification and classification results of the current lubrication state, the lubrication state is analyzed in time series, and the lubrication state value under the lubrication state is predicted by calculation to give an early warning, including the following steps: The current lubrication state is identified based on the trained deep belief network model; by continuously monitoring the lubrication system of the large machine, whether there is a transition from one lubrication state to another is identified; When there is a transition, the monitoring data of the lubrication status of the large machine at different time points are collected, and the collected monitoring data are arranged in chronological order to form a time series data set; A long short-term memory network model is constructed, and the time series data set is input into the long short-term memory network model for training; the lubrication state of the next time step is predicted by inputting the monitoring data of the current lubrication state identified by the deep belief network model into the trained long short-term memory network model; The sliding window method is used to gradually predict the lubrication status of multiple time steps in the future. Starting from the current window of real-time monitoring, the long short-term memory network model is used to predict the lubrication status of the next time step; by sliding the window forward one time step, adding the latest real-time monitoring data, removing the earliest monitoring data, and repeating the prediction process; by continuously sliding the window and using the long short-term memory network model for prediction, the lubrication status of multiple time steps in the future is gradually obtained; For the predicted lubrication state of multiple time steps in the future, the monitoring data under the predicted lubrication state is obtained; the lubrication state value formula under the predicted lubrication state is calculated: , in, Indicates the lubrication status value, represents the weight of the i-th sensor, represents the monitoring value of the i-th sensor at the j-th time step, represents the minimum value of the historical monitoring value of the i-th sensor, represents the maximum value of the historical monitoring value of the i-th sensor; t is the length of the time window; A lubrication status life end threshold is set. When the lubrication status value is lower than the threshold, the life end is reached and the early warning mechanism is triggered to issue an early warning.

5. According to the Internet of Things-based real-time monitoring and early warning system for lubrication status of large machines according to claim 1, it is characterized in that: Combining digital twin technology, a digital twin model of the lubrication status of the large machine is constructed to simulate the lubrication status of the large machine, including the following steps: According to the actual geometric shape and size of the large machine lubrication system, a geometric model is constructed using 3D modeling software, which describes the various components of the lubrication system, including: the shape, position and interconnection relationship of the bearings, gears, oil tanks, oil pumps, oil pipes, and lubrication points; Based on the constructed geometric model, the flow state of lubricant in the large machine lubrication system is simulated by solving the fluid dynamics equation, including oil pressure distribution and flow velocity distribution; the temperature distribution of lubricant in the large machine lubrication system and the heat transfer process are obtained by solving the heat conduction equation; the distribution of oil pressure, oil film thickness and friction parameters in the lubrication system is obtained by solving the mechanical equilibrium equation; By integrating the above geometric models with the results of fluid dynamics, heat conduction and mechanical balance analysis, a digital twin model of the lubrication status of the large machine was constructed; The monitoring data of the large machine lubrication system is collected in real time through IoT sensors, and the digital twin model is associated with the monitoring data collected by the IoT sensors. The actual operating status of the large machine lubrication system in the digital twin model is synchronized in real time through the data interface; The trained deep belief network model is deployed into the digital twin model, the current real-time lubrication status is output through the deep belief network model, and simulated and displayed through the digital twin model.

6. The real-time monitoring and early warning system for lubrication status of large machines based on the Internet of Things according to claim 1 is characterized in that: Establishing a fault identification model based on monitoring data includes the following steps: The fault recognition model is established by combining the convolutional neural network model in deep learning with the recurrent neural network model; The historical monitoring data set of the lubrication status of the large machine is collected and processed, and the fault types are marked, including leakage and blockage; the annotated historical monitoring data set is input into the fault recognition model for training; The monitoring data collected and processed through real-time monitoring of the Internet of Things is input into the trained fault identification model, which outputs whether a fault is identified under the current lubrication state of the large machine and the type of fault.

7. The real-time monitoring and early warning system for lubrication status of large machines based on the Internet of Things according to claim 6 is characterized in that: According to the fault identification results combined with the digital twin model, the flow path of the lubricant is simulated, and an early warning is issued by calculating the influence coefficient of the fault point on the lubricant flow, including the following steps: Deploy the trained fault identification model to the digital twin model, and identify the fault points in the lubricant flow path, including leakage points and blockage points, based on the fault identification model combined with the digital twin model. According to the identification results, count the total number K of all fault points in the flow path. The flow path of the lubricant in the large machine equipment is simulated according to the digital twin model and divided into a normal flow path and an abnormal propagation path; the abnormal propagation path is the flow path of the large machine lubrication state output by the fault identification model under the fault; During the simulation, computational fluid dynamics methods are used to collect the flow conditions of the lubricant on the normal flow path and the abnormal propagation path, including: flow velocity, temperature and pressure distribution; The influence coefficient formula of the fault point on the lubricant flow is calculated by: , in, is the influence coefficient, is the pressure under the kth fault point, is the flow rate at the kth fault point, is the temperature at the kth fault point, is the pressure of the normal flow path, is the flow velocity of the normal flow path, is the temperature of the normal flow path, , and are the rates of change of pressure, flow rate and temperature at the kth fault point over time, respectively, and K represents the total number of all fault points on the flow path; when the influence coefficient exceeds the threshold of 0.2, the early warning mechanism is triggered to issue an early warning.

8. The real-time monitoring and early warning system for lubrication status of large machines based on the Internet of Things according to claim 1 is characterized in that: The digital twin model is used to visualize the warning information, including the following steps: based on the calculated results of the lubrication state value and the influence coefficient of the fault point on the lubricant flow under the predicted lubrication state, when the warning mechanism is triggered, the generated warning information is integrated into the digital twin model, and the warning information includes: warning type, warning location and warning time.

Images