Real-time intelligent prediction method for fatigue life of mine hoist based on digital twinning

Abstract

The invention discloses a real-time intelligent prediction method for fatigue life of a mine hoist based on digital twinning. A physical scale model is established; establishing a digital twinborn model; aggregating the multi-source monitoring data and the elevator use information from the physical scale model and the simulation data from the elevator finite element model to obtain original aggregated data, and extracting key features by adopting a feature selection model; the method comprises the following steps of: predicting the service life of an entity elevator, training an elevator service life prediction model by using historical characteristic data to obtain a trained elevator service life prediction model, receiving multi-source monitoring data from a physical scale model in real time, predicting the service life of the elevator, and obtaining a prediction result and a corresponding adjustment control decision. And then an adjustment control decision is carried out on the finite element model of the elevator, the prediction precision is adjusted according to a simulation result, and finally a real-time intelligent prediction model of the fatigue life of the elevator is obtained. The method can effectively predict the fatigue life of the mine hoist in real time.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent monitoring and prediction, and specifically is a real-time intelligent prediction method for fatigue life of a mine hoist based on digital twins. Background Art

[0002] As an indispensable part of modern construction projects, hoists are also the core transportation equipment of the auxiliary transportation system of coal mines. They play a key role in the efficient flow of personnel and materials within coal mines. The stable operation of hoists is directly related to the safety and efficiency of coal mine production. However, under long-term high-frequency and variable load operation, harsh working environment and potential improper operation and maintenance conditions, the structural components of the hoist are prone to fatigue damage. During long-term operation, the damage will gradually accumulate, eventually leading to the degradation of the structure and function of the hoist, which will significantly reduce the carrying capacity of the hoist and make it unable to continue to support the predetermined load. This structural and functional degradation will not only reduce operating efficiency, but may also become a hidden danger that induces major safety accidents. Real-time and accurate prediction of the fatigue life of the hoist is an important measure to prevent a series of safety risks caused by equipment aging and ensure the safe and orderly operation of coal mines. It has far-reaching practical significance.

[0003] At present, although there are many methods for evaluating the fatigue life of hoists, such as the traditional methods based on SN curve (stress-life curve) and fracture mechanics, and the evaluation models based on probability theory and statistics, most of these methods rely on simplified mechanical assumptions and data under laboratory conditions, and it is difficult to fully capture and reflect the complex load changes and environmental fluctuations encountered by mine hoists in actual operations. In addition, hoists with different structural designs and material selections have different fatigue characteristics and influencing factors, which further increases the difficulty and complexity of fatigue life prediction.

[0004] In recent years, with the rapid development of global information technology, digital twin technology, as an emerging force, is gradually becoming a key driving force for the digital transformation of various industries. This technology achieves the deep integration and real-time interaction of physical space and virtual space by creating highly realistic virtual images for physical entities. In the field of fatigue life prediction of mine hoists, the application of digital twin technology means that various state changes of hoists during operation can be captured and simulated in real time, including but not limited to load distribution, stress response, ambient temperature and humidity, etc., thereby providing a more accurate and comprehensive data basis for fatigue life analysis.

[0005] Through digital twin technology, not only can the current state of the hoist be monitored and evaluated in real time, but also the fatigue damage and life changes that may occur in the future can be predicted by simulating the operation scenarios under different working conditions, providing a scientific basis for decision makers to achieve preventive maintenance and avoid production interruptions and safety accidents caused by equipment failure. Therefore, the application of digital twin technology to the fatigue life prediction of mine hoists not only has significant technical advantages, but also shows broad application prospects, and is expected to contribute to promoting the improvement of the safety management level of mine hoists in my country and ensuring the safety of coal mine production. Summary of the invention

[0006] In response to the problems existing in the above-mentioned prior art, the present invention provides a real-time intelligent prediction method for the fatigue life of a mine hoist based on digital twins. The method can make real-time and effective predictions on the fatigue life of a mine hoist, and can effectively monitor the failure risk of the mine hoist caused by excessive use time, thereby effectively avoiding the occurrence probability of safety accidents, and ensuring the safe and efficient production operations. At the same time, it is conducive to the realization of intelligent regulation and management of mine hoists, and helps to extend the service life of the hoist.

[0007] In order to achieve the above object, the present invention provides a real-time intelligent prediction method for fatigue life of a mine hoist based on digital twin, comprising the following steps:

[0008] Step 1: Establish a physical scale model and simulate the damaged operating conditions;

[0009] S11: Based on the similarity principle, a physical scale model is constructed in proportion according to formula (1) combined with the geometric properties of the components of the physical mine hoist;

[0010]

[0011] In the formula, S g is the geometric similarity factor; S is the physical scale model; R is the physical mine hoist; L, W, H are the length, height and width data respectively;

[0012] S12: configuring an intelligent monitoring component for the physical scale model, the intelligent monitoring component comprising a data acquisition module, multiple sensors, a microprocessor, a data storage module and a wireless communication module 1, wherein the multiple sensors are respectively installed at multiple nodes of the physical scale model and connected to the data acquisition module, and the microprocessor is respectively connected to the data acquisition module, the data storage module and the wireless communication module 1;

[0013] S13: applying a variable amplitude load to the physical scale model, simulating a damage accumulation cyclic loading condition using the physical scale model, synchronously using a variety of sensors to collect multi-source monitoring signals in real time, and sending them to a data acquisition module, the data acquisition module attaches a timestamp after receiving the multi-source monitoring signal, and then sends the multi-source monitoring signal with the timestamp to a microprocessor, the microprocessor obtains multi-source monitoring data according to the multi-source monitoring signal with the timestamp, performs noise reduction processing on it, obtains the multi-source monitoring data after noise reduction, and stores it in a data storage module;

[0014] While acquiring the multi-source monitoring data, synchronously acquire the use information of the hoist, and store the use information of the hoist in the data storage module, wherein the use information of the hoist at least includes the wire rope diameter and drum radius data;

[0015] Step 2: Establish a digital twin model and simulate the damage accumulation cyclic loading condition;

[0016] S21: Based on the digital twin technology, according to the multi-source monitoring data from the physical scale model and the use information of the elevator, combined with the connection characteristics of the components of the physical scale model and the working parameters under normal working conditions, a finite element model of the elevator, that is, a virtual simulation model of the elevator, is established on an industrial computer;

[0017] S22: Use ABAQUS software to perform nonlinear static and dynamic analysis on the physical scale model, accurately replicate the lifting operation of the physical scale model, and use the damage accumulation method to simulate the damage accumulation cyclic loading condition through the virtual simulation model of the hoist, and obtain simulation data in real time;

[0018] The simulation process of the damage accumulation cyclic loading condition is as follows:

[0019] S22-1: Setting the damage operation parameters for applying variable amplitude load;

[0020] S22-2: simulating the damage movement of the elevator virtual simulation model;

[0021] S23-3: Update the damage index field output in real time to understand the damage status in time;

[0022] S22-4: Execute S24-1 to S24-3 repeatedly until the set number of cycles N is reached and the process ends;

[0023] Step 3: Collect and process data from the finite element model and physical scale model of the hoist using an IoT-based connection system;

[0024] S31: constructing a connection system based on the Internet of Things, wherein the connection system based on the Internet of Things includes a wireless communication module 2, a data processor, and a database, wherein the data processor is connected to the wireless communication module 2 and the database respectively;

[0025] S32: establishing a wireless communication link 1 between the wireless communication module 1 and the wireless communication module 2, establishing a wireless communication link 2 between the communication module on the industrial computer and the wireless communication module 2, and realizing a two-way communication connection between the connection system based on the Internet of Things and the intelligent monitoring component and the hoist finite element model;

[0026] S33: during the simulation of the damage accumulation cyclic loading condition by the physical scale model and the simulation operation of the damage accumulation cyclic loading condition by the elevator virtual simulation model, a connection system based on the Internet of Things is used to receive multi-source monitoring data and elevator usage information from the physical scale model in real time, and to receive simulation data from the elevator virtual simulation model in real time. After receiving the multi-source monitoring data and simulation data, the data processor updates the timestamp to ensure the real-time consistency of the multi-source monitoring data and simulation data.

[0027] S34: The data processor first fuses the multi-source monitoring data and the simulation data according to the timestamp, and then obtains the original aggregated data by combining the use information of the elevator, and then stores the original aggregated data in the original data storage area of ​​the database. At the same time, the data processor extracts key features from the original aggregated data to obtain feature data, and then labels the feature data through experts, and then stores the labeled feature data in the feature data storage area of ​​the database as historical feature data;

[0028] The process of data processor extracting key features from raw aggregated data is as follows:

[0029] S34-1: The data processor reads the historical original aggregated data stored in the database, and divides the historical original aggregated data into a training set and a test set in proportion;

[0030] S34-2: Use the random forest model as the feature selection model, use the training set to train the feature selection model, use different nestimor hyperparameters to select key features during the training process, and obtain the trained feature selection model after the training is completed;

[0031] S34-3: After the data processor receives the new raw aggregated data, it uses the feature selection model to extract key features and obtain feature data;

[0032] Step 4: Predict the life of the physical elevator;

[0033] S41: Use the support vector machine model as the elevator life prediction model, and use the HingeLoss function in formula (2) as the loss function L(y);

[0034] L(y)=max(0,1-t·y) (2);

[0035] In the formula, y is the prediction result of the model, and t is the true label of the sample;

[0036] S42: reading historical feature data from a database, and dividing the historical feature data into a training set and a test set in proportion, inputting the feature data in the training set into a life prediction model for an elevator, and predicting the life of the elevator through the life prediction model for an elevator. Meanwhile, during the training process, the loss function is minimized. After the training is completed, a trained life prediction model for an elevator is obtained;

[0037] S43: receiving multi-source monitoring data from the physical scale model in real time, and extracting key features using a feature selection model to obtain feature data, and then inputting the feature data as input data into a life prediction model for the elevator, performing life prediction using the elevator life prediction model, and outputting prediction results and corresponding adjustment control decisions;

[0038] The obtained adjustment control decision is implemented in the elevator virtual simulation model, and the prediction accuracy of the elevator life prediction model is verified according to the simulation results. When the prediction accuracy is lower than the set value, the parameters of the elevator life prediction model are adjusted until the prediction accuracy is higher than or equal to the set value, and the real-time intelligent prediction model of the elevator fatigue life is obtained;

[0039] S44: Using the monitoring sensor group installed on the physical mine hoist, the physical hoist monitoring data is collected in real time, and the hoist usage information of the physical mine hoist is input. The feature selection model is then used to extract key features to obtain the physical hoist feature data. Next, the physical hoist feature data is input as input data into the real-time intelligent prediction model for the fatigue life of the hoist. The real-time intelligent prediction model for the fatigue life of the hoist is used to predict the life of the physical hoist, and the prediction results and corresponding adjustment control decisions are output. At the same time, prediction charts and reports are generated based on the prediction results.

[0040] As a preferred embodiment, in step four S44, after the adjustment control decision is obtained, the adjustment control decision is implemented using the elevator virtual simulation model, the adjustment control decision is verified according to the simulation results, and a more scientific basis is provided to the decision maker according to the verification results.

[0041] As a preference, in step 1 S12, the multiple sensors include a weighing sensor and a deflection sensor, the weighing sensor is used to collect load signals, and the deflection sensor is used to collect deflection signals.

[0042] As a preference, in step 1 S13, the elevator usage information record includes static data, dynamic data, specification data, historical data, environmental data and comprehensive data.

[0043] As a preference, in step three S31, the database is a Firebase database.

[0044] Preferably, in step three S34-2, the key features include load, deflection, wire rope diameter, drum radius and fatigue damage rate.

[0045] As a preferred embodiment, in step 4 S44, the fatigue damage rate is obtained according to formula (3);

[0046]

[0047] In the formula, σ M is the maximum stress, is the mean stress, D is the current damage state; Δσ is the stress range within one cycle; N is the number of cycles;

[0048] As a preferred embodiment, in S34-1 of step three, the original aggregated data is divided into a training set and a test set in a ratio of 7:3.

[0049] As a preferred embodiment, in step 4 S42, the historical feature data is divided into a training set and a test set in a ratio of 7:3.

[0050] In the present invention, a physical scale model is used to replace the physical mine hoist according to the similar principle, which can effectively solve the technical problem that the monitoring data of the physical hoist is difficult to obtain. Therefore, enough monitoring data can be collected conveniently, which is not only conducive to the subsequent construction process of the hoist virtual simulation model, but also can help to build a feature selection model with better classification effect and a hoist life prediction model with higher prediction accuracy. The hoist virtual simulation model is constructed by digital twin technology, and the physical scale model is analyzed nonlinearly by ABAQUS software to accurately copy the lifting operation of the physical scale model. Not only can the running state of the physical scale model be accurately simulated, but also when the physical scale model simulates the damage accumulation cycle loading condition, the damage accumulation cycle loading condition can be simulated by the hoist virtual simulation model synchronously. In this way, multi-source monitoring data from the physical scale model and simulation data from the hoist virtual simulation model can be obtained synchronously, which effectively makes up for the shortcomings and limitations of the single data from the physical scale model, and thus can ensure the training of a hoist life prediction model with better prediction accuracy. During the simulation operation, the damage accumulation method can accurately simulate the fatigue damage of the hoist during the variable amplitude fatigue loading process, and thus provide more accurate sample data for the training of the hoist life prediction model. For the trained hoist life prediction model, it is not directly applied to the life prediction of the physical mine hoist, but first receives it from the physical scale model in real time and performs online life prediction. Then the prediction results and the corresponding adjustment control decisions are first applied to the hoist virtual simulation model, and the prediction accuracy is verified by simulation. When the prediction accuracy cannot meet the requirements, the parameters of the model are adjusted to make the prediction accuracy meet the requirements, and finally the real-time intelligent prediction model of the hoist fatigue life is obtained, which can effectively ensure the prediction accuracy and prediction effect of the real-time intelligent prediction model of the hoist fatigue life. The real-time intelligent prediction model of the hoist fatigue life obtained after this adjustment can be effectively applied to the life prediction of the physical mine hoist, and can obtain effective adjustment control decisions. The real-time intelligent prediction model of the fatigue life of the hoist can predict the current life in real time according to the actual operating status of the mine hoist, and can display the prediction results to the operator or administrator in the form of charts and reports. By viewing these charts and reports, the operator or administrator can intuitively understand the service life of the mine hoist and the changes in the life of the hoist under different working conditions and loads. At the same time, based on the prediction information and decision-making suggestions provided by the hoist life prediction model, the operator or administrator can directly implement the optimization and adjustment plan according to the decision-making suggestions, and then flexibly adjust the load and other operating conditions to optimize the use efficiency of the hoist and extend its service life. At the same time, after connecting the real-time intelligent prediction model of the fatigue life of the hoist to the control system of the physical mine hoist, the physical mine hoist can be monitored and analyzed in real time, so as to realize active decision-making and feedback loop regulation operations.

[0051] The intelligent prediction method can not only monitor and evaluate the current state of the hoist in real time, but also effectively use the hoist virtual simulation model to simulate the operation scenarios under different working conditions, predict the fatigue damage and life changes that may occur in the future, and provide decision makers with a more scientific basis for adjustment and control. In this way, preventive maintenance can be achieved to avoid production interruptions and safety accidents caused by equipment failure. Therefore, the application of digital twin technology to fatigue life prediction of mine hoists not only has significant technical advantages, but also shows broad application prospects, and is expected to contribute to promoting the improvement of mine hoist safety management level and ensuring coal mine production safety.

[0052] This method can make real-time and effective predictions on the fatigue life of mine hoists, and can effectively monitor the failure risk of mine hoists due to excessive use, thereby effectively avoiding the probability of safety accidents and ensuring the safe and efficient production operations. At the same time, it is conducive to the realization of intelligent regulation and management of mine hoists, which helps to extend the service life of hoists. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a flow chart of the present invention;

[0054] Figure 2 A flow chart of the fatigue damage accumulation operation of the hoist in the present invention;

[0055] Figure 3 It is a five-layer structure and data flow diagram of the Internet of Things system in the present invention;

[0056] Figure 4 This is the architecture diagram of the random forest model in the present invention. DETAILED DESCRIPTION

[0057] The present invention will be further described below in conjunction with the accompanying drawings.

[0058] like Figure 1 and Figure 2 As shown, the present invention provides a real-time intelligent prediction method for fatigue life of a mine hoist based on digital twin, comprising the following steps:

[0059] Step 1: Establish a physical scale model and simulate the damaged operating conditions;

[0060] S11: Based on the similarity principle, according to formula (1) and the geometric properties of each component of the physical mine hoist, a physical scale model is constructed in proportion, and the physical scale model includes the geometric properties of all components of the physical mine hoist;

[0061]

[0062] In the formula, S g is the geometric similarity factor; S is the physical scale model; R is the physical mine hoist; L, W, H are the length, height and width data respectively;

[0063] Since it is difficult to collect monitoring data of the physical mine hoist, and it is also difficult to operate and control it during experimental testing, using a physical scale model instead of the physical mine hoist can conveniently collect relevant monitoring data, and then facilitate the analysis of the physical mine hoist according to the similarity principle.

[0064] S12: configuring an intelligent monitoring component for the physical scale model, the intelligent monitoring component comprising a data acquisition module, multiple sensors, a microprocessor, a data storage module and a wireless communication module 1, wherein the multiple sensors are respectively installed at multiple nodes of the physical scale model and connected to the data acquisition module, and the microprocessor is respectively connected to the data acquisition module, the data storage module and the wireless communication module 1;

[0065] Installing multiple sensors on the physical scale model can facilitate real-time collection of multi-source monitoring data, and then during the operation of the physical scale model, various parameters of the operation process can be accurately captured, so as to provide reliable technical support for subsequent fatigue life prediction; as a preferred embodiment, the multiple sensors include weighing sensors and deflection sensors, the weighing sensors are used to collect load signals, and the deflection sensors are used to collect deflection signals, and the microprocessor can obtain load measurement data and deflection measurement data based on the load signals and deflection signals.

[0066] S13: applying a variable amplitude load to the physical scale model, simulating a damage accumulation cyclic loading condition using the physical scale model, synchronously using a variety of sensors to collect multi-source monitoring signals in real time, and sending them to a data acquisition module, the data acquisition module attaches a timestamp after receiving the multi-source monitoring signal, and then sends the multi-source monitoring signal with the timestamp to a microprocessor, the microprocessor obtains multi-source monitoring data according to the multi-source monitoring signal with the timestamp, performs noise reduction processing on it, obtains the multi-source monitoring data after noise reduction, and stores it in a data storage module;

[0067] While acquiring multi-source monitoring data, the use information of the hoist is acquired synchronously and stored in the data storage module, wherein the use information of the hoist includes at least the wire rope diameter and drum radius data; the data acquisition module is used to add a timestamp to the multi-source monitoring signal, which can effectively record the specific time when each monitoring data is generated, ensuring the real-time and synchronization of the data, and can facilitate the analysis of the operating status of the hoist at different time points;

[0068] Step 2: Establish a digital twin model and simulate the damage accumulation cyclic loading condition;

[0069] S21: Based on the digital twin technology, according to the multi-source monitoring data from the physical scale model and the use information of the elevator, and combined with the connection characteristics of the components of the physical scale model and the working parameters under normal working conditions, a finite element model of the elevator, that is, a virtual simulation model of the elevator, is established on an industrial computer; the constructed virtual simulation model of the elevator is used as a simulation platform for analyzing the operation and structure of the elevator;

[0070] S22: Use ABAQUS software to perform nonlinear static and dynamic analysis on the physical scale model, accurately replicate the lifting operation of the physical scale model, and use the damage accumulation method to simulate the damage accumulation cyclic loading condition through the virtual simulation model of the hoist, and obtain simulation data in real time;

[0071] Finite element analysis can be performed under various loads, including lifting loads, static loads and dynamic loads. When performing nonlinear static analysis of the hoist structure, both material nonlinearity and geometric nonlinearity are considered, which can provide a more accurate prediction of the structural response. As a preference, when performing finite element analysis, the working conditions of the physical scale model are considered at the same time, and the static, dynamic and moving masses on the hoist are fully considered. In addition, all factors related to the operation and life of the hoist are included in the hoist virtual simulation model as real-time external data.

[0072] As a preferred embodiment, the specific process of finite element analysis is as follows:

[0073] S21-1: Use computer-aided design (CAD) software to create a three-dimensional geometric model of the physical scale model, and the three-dimensional geometric model should accurately capture the characteristics, geometry, material properties, behavior, physical rules, functions and operating environment of the physical scale model; as a preferred embodiment, it is necessary to integrate the use information of the elevator for simulation and prediction of the virtual simulation model of the elevator, and at the same time, in the process of creating the three-dimensional geometric model, it is also possible to accurately construct a virtual space that is compatible with the physical scale model;

[0074] S21-2: Define the physical and mechanical properties of the constituent materials of the physical scale model. These properties can directly affect the accuracy and reliability of the finite element analysis.

[0075] S21-3: Divide the three-dimensional geometric model into multiple small finite element units for numerical calculation. The density and quality of the mesh directly affect the analysis results and calculation time. Critical areas (such as stress concentration areas) need to be set with dense meshes.

[0076] S21-4: Define the boundary conditions of the 3D geometric model, including support conditions, constraint conditions, etc. At the same time, define various load conditions according to the working conditions of the elevator, including static load, dynamic load, etc.;

[0077] S21-5: Use finite element software ABAQUS to solve the three-dimensional geometric model and obtain analysis results to calculate the stress, strain, displacement and other responses of the hoist under various working conditions. Through subsequent processing tools, these structures can be viewed and analyzed, and then compared and verified based on the analysis results and actual test data or engineering experience;

[0078] The simulation process of the damage accumulation cyclic loading condition is as follows:

[0079] S22-1: Setting the damage operation parameters for applying variable amplitude load;

[0080] S22-2: simulating the damage movement of the elevator virtual simulation model;

[0081] S23-3: Update the damage index field output in real time to understand the damage status in time;

[0082] S22-4: Execute S24-1 to S24-3 repeatedly until the set number of cycles N is reached and the process ends;

[0083] The data obtained by finite element analysis can be used to facilitate the assessment of fatigue damage. There are two main methods for assessing fatigue damage: (1) constant amplitude fatigue loading; (2) variable amplitude fatigue loading. The loads experienced by the hoist are static and cyclical, resulting in fatigue damage accumulation that varies with each loading cycle. Therefore, the traditional linear cumulative damage is not applicable to the hoist case. In the damage analysis of mine hoist structures, a variety of parameters can be used to assess the damage, including total strain, residual stress, stiffness degradation, strength degradation, heat dissipation caused by microcracks, crack propagation, and sound velocity. By considering stress, displacement, and damage index, and applying variable amplitude moving loads, the mechanical response of the hoist structure is simulated. During variable amplitude fatigue loading, the hoist structure will undergo irreversible deformation and accumulate residual stress, which will cause internal cracks in the hoist structure, seriously affecting the overall stability and reliability of the hoist. By combining the fatigue loss rate of the hoist, the real-time perception data transmitted by the sensor (including load data, deflection acuity), and the geometric parameter data (wire rope diameter, drum radius), the fatigue life of the hoist can be accurately assessed.

[0084] Step 3: Collect and process data from the finite element model and physical scale model of the hoist using an IoT-based connection system;

[0085] S31: constructing a connection system based on the Internet of Things, wherein the connection system based on the Internet of Things includes a wireless communication module 2, a data processor, and a database, wherein the data processor is connected to the wireless communication module 2 and the database respectively;

[0086] S32: Establish wireless communication link one between wireless communication module one and wireless communication module two, establish wireless communication link two between the communication module on the industrial computer and wireless communication module two, and realize two-way communication connection between the connection system based on the Internet of Things and the intelligent monitoring component and the finite element model of the elevator; thereby, not only can the real-time status of the physical scale model be monitored by the connection system based on the Internet of Things, but also the simulation data of the finite element model of the elevator can be obtained by the connection system based on the Internet of Things; the connection system based on the Internet of Things realizes a two-way connection between the physical space (physical scale model and intelligent monitoring component) and the virtual space (finite element model of the elevator), and this connection ensures real-time synchronization of data, so that the physical space and the virtual space can be seamlessly connected, the actual operating status of the elevator can be reflected in time, and the data in the physical space and the virtual space can be obtained synchronously.

[0087] The physical scale model, intelligent monitoring components and IoT-based connection system constitute the IoT system. In this prediction method, the IoT plays an important role as the source of real-time data and the source of the mechanism for transmitting data. The IoT system has a five-layer structure, namely, the object layer, the sensor layer, the communication layer, the data management layer and the application layer. Figure 3 As shown; the object layer is mainly composed of physical scale models; the sensor layer is mainly composed of various sensors in the intelligent monitoring components, such as weighing sensors and deflection sensors, which are mainly used for data collection; the communication layer mainly includes wireless communication modules one and two, which are used to collect and transmit data, including serial port converters and communication modules, which promote the seamless flow and reliable transmission of data within the Internet of Things system; the data management layer is mainly composed of databases and data storage modules, which are used to store and integrate basic data converted into different formats. The application layer is the basic component of the Internet of Things system, mainly composed of microprocessors and data processors, which mainly perform data analysis and processing, including data association, data mining and data analysis;

[0088] S33: During the simulation of the damage accumulation cyclic loading condition by the physical scale model and the simulation operation of the damage accumulation cyclic loading condition by the elevator virtual simulation model, a connection system based on the Internet of Things is used to receive multi-source monitoring data and elevator usage information from the physical scale model in real time, and to receive simulation data from the elevator virtual simulation model in real time. After receiving the multi-source monitoring data and simulation data, the data processing updates the timestamp, thereby ensuring the real-time consistency and accuracy of the multi-source monitoring data and simulation data in the subsequent aggregation process;

[0089] S34: The data processor first fuses the multi-source monitoring data and simulation data according to the timestamp, and then obtains the original aggregated data in combination with the use information of the elevator, and then stores the original aggregated data in the original data storage area of ​​the database. At the same time, the data processor extracts key features of the original aggregated data to obtain feature data, and then labels the feature data through experts, and then stores the labeled feature data in the feature data storage area of ​​the database as historical feature data; first fusing the data and then extracting the key features can effectively extract the key features of the monitoring data in the physical scale model and the key features of the monitoring data in the elevator virtual simulation model, which can effectively solve the complexity brought by data heterogeneity and fully obtain various factors affecting the fatigue life of the elevator;

[0090] The process of data processor extracting key features from raw aggregated data is as follows:

[0091] S34-1: The data processor reads the historical original aggregated data stored in the database, and divides the historical original aggregated data into a training set and a test set in proportion;

[0092] S34-2: Figure 4 As shown, a random forest (RF) model is used as a feature selection model, and the feature selection model is trained using a training set. Different nestimor hyperparameters are used to select key features during the training process, and a trained feature selection model is obtained after the training is completed;

[0093] The random forest (RF) model combines the CART tree and random subspace, where the CART tree consists of decision nodes, leaf nodes, and root nodes. Each tree is built using an independent random sample method, ensuring the diversity and stability of the model. Subsets are created through the bagging technique, and random subsets of feature attributes are selected according to specific criteria, further increasing the generalization ability of the model.

[0094] The most commonly used combination strategy for random forest models is the voting method, which is calculated as follows:

[0095] in, Indicates h i In category c j Output on;

[0096] Through the analysis of the random forest model, the importance values ​​of load, deflection, wire rope diameter, drum radius and fatigue damage rate in the fatigue life prediction process are 0.37, 0.34, 0.13, 0.0 and 0.2 respectively.

[0097] S34-3: After the data processor receives the new raw aggregated data, it uses the feature selection model to extract key features and obtain feature data;

[0098] Step 4: Predict the life of the physical elevator;

[0099] S41: A similarity-based one-class classification method is adopted, and the support vector machine (SVM) model is used as the elevator life prediction model, and the Hinge Loss function in formula (2) is used as the loss function L(y); when the loss value is extended to the entire training data set, the loss is

[0100] L(y)=max(0,1-t·y) (2);

[0101] In the formula, y is the prediction result of the model, and t is the true label of the sample;

[0102] Due to the complexity of multiple variables, it is challenging to predict the life of the elevator using traditional mathematical models. In terms of processing the intrinsic connection of data, machine learning (ML) methods can make more accurate predictions. Since the SVM model is good at processing high-dimensional data with obvious edges and scattered data points, the SVM model is used to predict the life of the elevator. The support vector machine (SVM) model learns the characteristics of typical scenarios and predicts whether the new input deviates from these specific situations. Finally, the accuracy of the model is analyzed through k-fold cross validation to verify the data set.

[0103] The input data of the hoist life prediction model include load, deflection, wire rope diameter, drum radius and fatigue damage rate, and the output is the hoist life prediction result. The evaluation of the hoist life prediction model is based on two criteria: mean square error (MSE) and determination coefficient (R 2 ), mean square error (MSE) refers to the square error between the predicted value and the actual value, reflecting the prediction error. 2 ) is used to quantify how well the model fits the data. The smaller the mean square error (MSE), the better the coefficient of determination (R 2 ) is closer to 1, indicating that the prediction error is smaller and the performance of the prediction model is better.

[0104] S42: reading historical feature data from a database, and dividing the historical feature data into a training set and a test set in proportion, inputting the feature data in the training set into the elevator life prediction model, and predicting the life of the elevator through the elevator life prediction model. Meanwhile, during the training process, the loss function is minimized. After the training is completed, a trained elevator life prediction model is obtained. Since the historical feature data is feature data with labels, the trained elevator life prediction model can directly output life prediction according to the input data.

[0105] S43: receiving multi-source monitoring data from the physical scale model in real time, and extracting key features using a feature selection model to obtain feature data, and then inputting the feature data as input data into a life prediction model for the elevator, performing life prediction using the elevator life prediction model, and outputting prediction results and corresponding adjustment control decisions;

[0106] The obtained adjustment control decision is implemented in the elevator virtual simulation model, and the prediction accuracy of the elevator life prediction model is verified according to the simulation results. When the prediction accuracy is lower than the set value, the parameters of the elevator life prediction model are adjusted until the prediction accuracy is higher than or equal to the set value, and the real-time intelligent prediction model of the elevator fatigue life is obtained;

[0107] S44: Using the monitoring sensor group installed on the physical mine hoist, the physical hoist monitoring data is collected in real time, and the hoist usage information of the physical mine hoist is input. The feature selection model is then used to extract key features to obtain the physical hoist feature data. Next, the physical hoist feature data is input as input data into the real-time intelligent prediction model for the fatigue life of the hoist. The real-time intelligent prediction model for the fatigue life of the hoist is used to predict the life of the physical hoist, and the prediction results and corresponding adjustment control decisions are output. At the same time, prediction charts and reports are generated based on the prediction results.

[0108] As a preferred embodiment, in step four S44, after the adjustment control decision is obtained, the adjustment control decision is implemented using the elevator virtual simulation model, the adjustment control decision is verified according to the simulation results, and a more scientific basis is provided to the decision maker according to the verification results.

[0109] Preferably, in step 1 S13, the hoist usage information record includes static data, dynamic data, specification data, historical data, environmental data and comprehensive data. Preferably, the static data includes geometric properties, attributes and materials, wherein the geometric properties include size, lifting height, lifting angle / inclination, track / guide rail specifications, the attributes include manufacturer, model, serial number, installation date, and place of use, and the materials include main structure material, rope / chain material, telegraph element material, wear-resistant / anti-corrosion coating; the dynamic data includes sensor data, built-in data, and operation data, wherein the sensor data includes load sensor readings, speed sensor readings, position sensor readings, and temperature sensor readings, the built-in data includes running time accumulation, fault records and diagnostic information, maintenance records, emergency stop times and reasons, and the operation data includes operation mode, control instruction records, and operator identity / number; the specification data The data includes standards and specifications, performance specifications, rules and restrictions, among which standards and specifications include safety standards to be followed, design specifications / standard numbers, environmental protection requirements and compliance; performance specifications include maximum lifting capacity, maximum lifting speed, motor power and type, brake system specifications; rules and restrictions include operating temperature range, allowable ambient temperature, regular inspection and maintenance cycle, and usage restrictions under special operating or environmental conditions; the historical data includes operation records, maintenance records, fault records, and repair records; the environmental data includes wind speed, temperature, humidity, and external working conditions, among which external working conditions include dust concentration, corrosive gas content, and electromagnetic interference; the comprehensive data includes statistical analysis reports, performance evaluation reports, technical documents, laws, regulations, and standard documents.

[0110] Preferably, in step 3 S31, the database is a Firebase database. Preferably, during the storage process, the data is converted into JSON format, which is a lightweight data exchange format. The JSON object consists of six layers of nodes, making the data more standardized and easy to process during transmission, which is beneficial for training machine learning (ML) modules because the sensor data features of different operating nodes are different. Therefore, the classification engine based on machine learning can effectively utilize these historical data.

[0111] Preferably, in step three S34-2, the key features include load, deflection, wire rope diameter, drum radius and fatigue damage rate.

[0112] As a preferred embodiment, the fatigue damage rate is obtained according to formula (3);

[0113]

[0114] In the formula, σ M is the maximum stress, is the mean stress, D is the current damage state; Δσ is the stress range within one cycle; N is the number of cycles;

[0115] As a preferred embodiment, in S34-1 of step three, the original aggregated data is divided into a training set and a test set in a ratio of 7:3.

[0116] As a preferred embodiment, in step 4 S42, the historical feature data is divided into a training set and a test set in a ratio of 7:3.

[0117] In the present invention, a physical scale model is used to replace the physical mine hoist according to the similar principle, which can effectively solve the technical problem that the monitoring data of the physical hoist is difficult to obtain. Therefore, enough monitoring data can be collected conveniently, which is not only conducive to the subsequent construction process of the hoist virtual simulation model, but also can help to build a feature selection model with better classification effect and a hoist life prediction model with higher prediction accuracy. The hoist virtual simulation model is constructed by digital twin technology, and the physical scale model is analyzed nonlinearly by ABAQUS software to accurately copy the lifting operation of the physical scale model. Not only can the running state of the physical scale model be accurately simulated, but also when the physical scale model simulates the damage accumulation cycle loading condition, the damage accumulation cycle loading condition can be simulated by the hoist virtual simulation model synchronously. In this way, multi-source monitoring data from the physical scale model and simulation data from the hoist virtual simulation model can be obtained synchronously, which effectively makes up for the shortcomings and limitations of the single data from the physical scale model, and thus can ensure the training of a hoist life prediction model with better prediction accuracy. During the simulation operation, the damage accumulation method can accurately simulate the fatigue damage of the hoist during the variable amplitude fatigue loading process, and thus provide more accurate sample data for the training of the hoist life prediction model. For the trained hoist life prediction model, it is not directly applied to the life prediction of the physical mine hoist, but first receives it from the physical scale model in real time and performs online life prediction. Then the prediction results and the corresponding adjustment control decisions are first applied to the hoist virtual simulation model, and the prediction accuracy is verified by simulation. When the prediction accuracy cannot meet the requirements, the parameters of the model are adjusted to make the prediction accuracy meet the requirements, and finally the real-time intelligent prediction model of the hoist fatigue life is obtained, which can effectively ensure the prediction accuracy and prediction effect of the real-time intelligent prediction model of the hoist fatigue life. The real-time intelligent prediction model of the hoist fatigue life obtained after this adjustment can be effectively applied to the life prediction of the physical mine hoist, and can obtain effective adjustment control decisions. The real-time intelligent prediction model of the fatigue life of the hoist can predict the current life in real time according to the actual operating status of the mine hoist, and can display the prediction results to the operator or administrator in the form of charts and reports. By viewing these charts and reports, the operator or administrator can intuitively understand the service life of the mine hoist and the changes in the life of the hoist under different working conditions and loads. At the same time, based on the prediction information and decision-making suggestions provided by the hoist life prediction model, the operator or administrator can directly implement the optimization and adjustment plan according to the decision-making suggestions, and then flexibly adjust the load and other operating conditions to optimize the use efficiency of the hoist and extend its service life. At the same time, after connecting the real-time intelligent prediction model of the fatigue life of the hoist to the control system of the physical mine hoist, the physical mine hoist can be monitored and analyzed in real time, so as to realize active decision-making and feedback loop regulation operations.

[0118] The intelligent prediction method can not only monitor and evaluate the current state of the hoist in real time, but also effectively use the hoist virtual simulation model to simulate the operation scenarios under different working conditions, predict the fatigue damage and life changes that may occur in the future, and provide decision makers with a more scientific basis for adjustment and control. In this way, preventive maintenance can be achieved to avoid production interruptions and safety accidents caused by equipment failure. Therefore, the application of digital twin technology to fatigue life prediction of mine hoists not only has significant technical advantages, but also shows broad application prospects, and is expected to contribute to promoting the improvement of mine hoist safety management level and ensuring coal mine production safety.

[0119] This method can make real-time and effective predictions on the fatigue life of mine hoists, and can effectively monitor the failure risk of mine hoists due to excessive use, thereby effectively avoiding the probability of safety accidents and ensuring the safe and efficient production operations. At the same time, it is conducive to the realization of intelligent regulation and management of mine hoists, which helps to extend the service life of hoists.

Claims

1. A real-time intelligent prediction method for fatigue life of mine hoist based on digital twin, characterized in that: The following steps are involved: Step 1: Establish a physical scale model and simulate the damaged operating conditions; S11: Based on the similarity principle, a physical scale model is constructed in proportion according to formula (1) combined with the geometric properties of the components of the physical mine hoist; In the formula, S g is the geometric similarity factor; S is a physical scale model; R is a physical mine hoist; L, W, and H are length, height, and width data respectively; S12: configuring an intelligent monitoring component for the physical scale model, the intelligent monitoring component comprising a data acquisition module, multiple sensors, a microprocessor, a data storage module and a wireless communication module 1, wherein the multiple sensors are respectively installed at multiple nodes of the physical scale model and connected to the data acquisition module, and the microprocessor is respectively connected to the data acquisition module, the data storage module and the wireless communication module 1; S13: applying a variable amplitude load to the physical scale model, simulating a damage accumulation cyclic loading condition using the physical scale model, synchronously using a variety of sensors to collect multi-source monitoring signals in real time, and sending them to a data acquisition module, the data acquisition module attaches a timestamp after receiving the multi-source monitoring signal, and then sends the multi-source monitoring signal with the timestamp to a microprocessor, the microprocessor obtains multi-source monitoring data according to the multi-source monitoring signal with the timestamp, performs noise reduction processing on it, obtains the multi-source monitoring data after noise reduction, and stores it in a data storage module; While acquiring the multi-source monitoring data, synchronously acquire the use information of the hoist, and store the use information of the hoist in the data storage module, wherein the use information of the hoist at least includes the wire rope diameter and drum radius data; Step 2: Establish a digital twin model and simulate the damage accumulation cyclic loading condition; S21: Based on the digital twin technology, according to the multi-source monitoring data from the physical scale model and the use information of the elevator, combined with the connection characteristics of the components of the physical scale model and the working parameters under normal working conditions, a finite element model of the elevator, that is, a virtual simulation model of the elevator, is established on an industrial computer; S22: Use ABAQUS software to perform nonlinear static and dynamic analysis on the physical scale model, accurately replicate the lifting operation of the physical scale model, and use the damage accumulation method to simulate the damage accumulation cyclic loading condition through the virtual simulation model of the hoist, and obtain simulation data in real time; The simulation process of the damage accumulation cyclic loading condition is as follows: S22-1: Setting the damage operation parameters for applying variable amplitude load; S22-2: simulating the damage movement of the elevator virtual simulation model; S23-3: Update the damage index field output in real time to understand the damage status in time; S22-4: Execute S24-1 to S24-3 repeatedly until the set number of cycles N is reached and the process ends; Step 3: Collect and process data from the finite element model and physical scale model of the hoist using an IoT-based connection system; S31: constructing a connection system based on the Internet of Things, wherein the connection system based on the Internet of Things includes a wireless communication module 2, a data processor, and a database, wherein the data processor is connected to the wireless communication module 2 and the database respectively; S32: establishing a wireless communication link 1 between the wireless communication module 1 and the wireless communication module 2, establishing a wireless communication link 2 between the communication module on the industrial computer and the wireless communication module 2, and realizing a two-way communication connection between the connection system based on the Internet of Things and the intelligent monitoring component and the hoist finite element model; S33: during the simulation of the damage accumulation cyclic loading condition by the physical scale model and the simulation operation of the damage accumulation cyclic loading condition by the elevator virtual simulation model, a connection system based on the Internet of Things is used to receive multi-source monitoring data and elevator usage information from the physical scale model in real time, and to receive simulation data from the elevator virtual simulation model in real time. After receiving the multi-source monitoring data and simulation data, the data processor updates the timestamp to ensure the real-time consistency of the multi-source monitoring data and simulation data. S34: The data processor first fuses the multi-source monitoring data and the simulation data according to the timestamp, and then obtains the original aggregated data by combining the use information of the elevator, and then stores the original aggregated data in the original data storage area of ​​the database. At the same time, the data processor extracts key features from the original aggregated data to obtain feature data, and then labels the feature data through experts, and then stores the labeled feature data in the feature data storage area of ​​the database as historical feature data; The process of data processor extracting key features from raw aggregated data is as follows: S34-1: The data processor reads the historical original aggregated data stored in the database, and divides the historical original aggregated data into a training set and a test set in proportion; S34-2: Use the random forest model as the feature selection model, use the training set to train the feature selection model, use different nestimor hyperparameters to select key features during the training process, and obtain the trained feature selection model after the training is completed; S34-3: After the data processor receives the new raw aggregated data, it uses the feature selection model to extract key features and obtain feature data; Step 4: Predict the life of the physical elevator; S41: Use the support vector machine model as the elevator life prediction model, and use the Hinge Loss function in formula (2) as the loss function L(y); L(y)=max(0,1-t·y) (2); In the formula, y is the prediction result of the model, and t is the true label of the sample; S42: reading historical feature data from a database, and dividing the historical feature data into a training set and a test set in proportion, inputting the feature data in the training set into a life prediction model for an elevator, and predicting the life of the elevator through the life prediction model for an elevator. Meanwhile, during the training process, the loss function is minimized. After the training is completed, a trained life prediction model for an elevator is obtained; S43: receiving multi-source monitoring data from the physical scale model in real time, and extracting key features using a feature selection model to obtain feature data, and then inputting the feature data as input data into a life prediction model for the elevator, performing life prediction using the elevator life prediction model, and outputting prediction results and corresponding adjustment control decisions; The obtained adjustment control decision is implemented in the elevator virtual simulation model, and the prediction accuracy of the elevator life prediction model is verified according to the simulation results. When the prediction accuracy is lower than the set value, the parameters of the elevator life prediction model are adjusted until the prediction accuracy is higher than or equal to the set value, and the real-time intelligent prediction model of the elevator fatigue life is obtained; S44: Using the monitoring sensor group installed on the physical mine hoist, the physical hoist monitoring data is collected in real time, and the hoist usage information of the physical mine hoist is input. The feature selection model is then used to extract key features to obtain the physical hoist feature data. Next, the physical hoist feature data is input as input data into the real-time intelligent prediction model for the fatigue life of the hoist. The real-time intelligent prediction model for the fatigue life of the hoist is used to predict the life of the physical hoist, and the prediction results and corresponding adjustment control decisions are output. At the same time, prediction charts and reports are generated based on the prediction results.

2. According to claim 1, a real-time intelligent prediction method for fatigue life of a mine hoist based on digital twins is characterized in that: In step 4 S44, after the adjustment control decision is obtained, the adjustment control decision is implemented using the elevator virtual simulation model, the adjustment control decision is verified according to the simulation results, and a more scientific basis is provided to the decision maker according to the verification results.

3. The real-time intelligent prediction method for fatigue life of a mine hoist based on digital twin according to claim 1 is characterized in that: In step 1 S12, the multiple sensors include a weighing sensor and a deflection sensor, the weighing sensor is used to collect a load signal, and the deflection sensor is used to collect a deflection signal.

4. The real-time intelligent prediction method for fatigue life of a mine hoist based on digital twin according to claim 1 is characterized in that: In step 1 S13, the elevator usage information record includes static data, dynamic data, specification data, historical data, environmental data and comprehensive data.

5. The real-time intelligent prediction method for fatigue life of a mine hoist based on digital twin according to claim 1 is characterized in that: In step three S31, the database is a Firebase database.

6. The real-time intelligent prediction method for fatigue life of a mine hoist based on digital twin according to claim 1 is characterized in that: In step three S34-2, the key features include load, deflection, wire rope diameter, drum radius and fatigue damage rate.

7. The real-time intelligent prediction method for fatigue life of a mine hoist based on digital twin according to claim 1 is characterized in that: In step 4 S44, the fatigue damage rate is obtained according to formula (3); In the formula, σ M is the maximum stress, is the mean stress, D is the current damage state; Δσ is the stress range within one cycle; N is the number of cycles.

8. The method for real-time intelligent prediction of fatigue life of mine hoist based on digital twin according to claim 1 is characterized in that: In step 3 S34-1, the original aggregated data is divided into a training set and a test set in a ratio of 7:

3.

9. The method for real-time intelligent prediction of fatigue life of mine hoist based on digital twin according to claim 1 is characterized in that: In step 4 S42, the historical feature data is divided into a training set and a test set in a ratio of 7:3.

Images