A downhole robot power unit digital twin monitoring method and system

Abstract

The application discloses a kind of underground robot power unit digital twin monitoring method and system, it is related to mine equipment intelligent monitoring and digital twin technical field.The method steps are:S1, the battery entity is carried out multimodal data acquisition;S2, the multimodal data is preprocessed and is transmitted into digital twin space;S3, the twin model of battery is constructed and the data imported is physically modeled;S4, the health state of battery is deduced, and future health state is predicted;S5, the gas concentration when underground robot works is monitored;S6, when the health state prediction value of battery and gas concentration are about to exceed safety threshold, early warning is prompted;S7, after early warning, the safety of underground robot is guaranteed by controlling the work of each safety element in underground work.The method fuses digital twin and multimodal big model technology, improves monitoring efficiency and accuracy, and provides reliable guarantee for underground robot operation safety.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent monitoring and digital twinning of mine equipment, in particular to a digital twinning monitoring method and system for a power unit of an underground robot. BACKGROUND

[0002] In the field of intelligent operation in underground space, real-time state monitoring of the power unit of an underground robot is a core technical link to ensure reliable operation of equipment and safety of operation. In the prior art, the power unit monitoring system generally uses a state of charge (SOC) estimation method based on an equivalent circuit model, combines a preset threshold alarm mechanism to realize basic early warning function, and relies on thermocouples or infrared sensors to build a limited temperature monitoring system. Although this kind of monitoring scheme can realize basic perception of part parameters of the power unit, it has fundamental defects in multi-physical field coupling analysis and real-time response capability.

[0003] The running process of the power unit of the underground robot is essentially a complex dynamic process of mutual coupling and synergistic action of multiple physical fields such as electric field, thermal field and mechanical field. However, the existing monitoring technology mostly adopts a single field independent analysis mode, lacking systematic modeling of the interaction mechanism of multiple physical fields. For example, in the process of battery charging and discharging, the heat generation effect caused by the change of electric field and the distribution of thermal field will in turn affect the electrochemical performance of the battery, but the traditional method does not establish an effective coupling model, resulting in significant one-sidedness and inaccuracy in the running state evaluation of the power unit, which cannot truly reflect the actual running conditions of the equipment.

[0004] At the same time, the existing monitoring system adopts a discrete sensor data acquisition architecture and a simple threshold determination logic, and the data processing flow includes multiple links such as acquisition, transmission, analysis and analysis, resulting in inherent delay in system response. When the power unit has abnormal conditions such as battery voltage drop and local overheating, the monitoring system cannot complete state recognition and early warning within a millisecond time scale, especially in the complex environment of underground high temperature, high humidity and strong electromagnetic interference. This response lag is extremely easy to cause delay in fault disposal, and further cause serious accidents such as battery thermal runaway and equipment shutdown, which seriously restricts the improvement of intelligent operation level of the underground robot.

[0005] With the development of underground space development to deeper and more complex areas, the deficiencies of the existing monitoring technology in multi-physical field synergistic analysis capability and real-time response efficiency have become a technical bottleneck hindering the safe and reliable operation of the underground robot, and new monitoring theories and technical schemes are urgently needed to meet the development needs of underground intelligent operation equipment. SUMMARY

[0006] The application aims to provide a downhole robot power unit digital twin monitoring method and system, which collects battery data through multiple sensors, constructs a multi-physical field coupling model, and fuses multi-modal large model analysis technology to realize efficient and accurate monitoring and trend prediction of the power unit state. At the same time, the gas sensor is used to ensure the safety of the working environment, and the early warning and emergency disposal module is combined to ensure the stable operation of the downhole robot under complex working conditions.

[0007] To achieve the above-mentioned purpose, the application provides a downhole robot power unit digital twin monitoring method, and the steps are as follows:

[0008] Step S1, multi-modal data acquisition of stress, temperature, electricity, deformation and thermal runaway related parameters of the battery;

[0009] Step S2, pre-processing the collected multi-modal data, and inputting the pre-processed data into the digital twin space;

[0010] Step S3, constructing a twin model of the battery in the twin space and physically modeling the input data to monitor the current battery state;

[0011] Step S4, inferring the health status of the battery through a multi-modal large model, and predicting the future health status of the battery;

[0012] Step S5, installing a gas sensor outside the downhole robot and in the explosion-proof cavity, respectively, to monitor the gas concentration when the downhole robot works;

[0013] Step S6, when the predicted value of the health status of the battery and the gas concentration are about to exceed the safety threshold, the early warning system in the twin space is used for early warning;

[0014] Step S7, after the early warning, the safety of the downhole robot working in the downhole is ensured by controlling the work of each safety element.

[0015] Preferably, in step S1, an ultrasonic sensor, a voltage and current detection module, a method based on a projection laser scanner, and a gas sensor are used to monitor and collect data of the downhole robot power unit, including:

[0016] Stress data and temperature data acquisition: an ultrasonic sensor is used to represent the change characteristics of the complex stress field inside the battery based on the ultrasonic elastic battery pack internal stress monitoring method; the ultrasonic sensor is composed of an ultrasonic longitudinal wave sensor and an ultrasonic surface wave sensor, and the ultrasonic longitudinal wave sensor is symmetrically installed on both sides of the battery to obtain the stress state inside the battery; two ultrasonic surface wave sensors are installed on the upper surface of the battery to obtain the surface tension state of the battery pack;

[0017] Electric parameter collection: the voltage and current detection module is used to collect the voltage and current inside the battery, the internal resistance value of the battery is calculated according to the collected data, and the battery capacity is indirectly judged by combining the battery temperature data;

[0018] Deformation data collection: a method based on projection laser scanning is used, a laser projection device and a CCD camera are respectively installed on the inner wall of the explosion-proof cavity above the battery, the optical axis of the CCD camera needs to be perpendicular to the reference plane, when the battery deforms due to complex working conditions, the CCD camera senses the deformation of the battery by collecting the projection stripes of the battery surface deformation;

[0019] Thermal runaway monitoring: a gas sensor is used to monitor the thermal runaway of the battery; the gas sensor is installed on the inner wall of the explosion-proof cavity, and the concentration of the gas generated by the thermal runaway of the battery is monitored to predict the early thermal runaway of the battery.

[0020] Preferably, in step S2, the collected multi-modal data is preprocessed, and the steps are as follows:

[0021] Step S21, decompose the signal into different frequency components to remove noise by wavelet transform, and then align different modal data according to the time stamp;

[0022] Step S22, identify and eliminate abnormal values caused by measurement errors, and use a multi-source data fusion method to fill in the data loss caused by the complex communication environment in the well;

[0023] Step S23, use the minimum-maximum normalization method to normalize the voltage, current, temperature and Gas concentration related data, use Z-score normalization method to normalize the data with approximate normal distribution of stress and strain, convert the data to a unified scale, and store the normalized data in a unified format;

[0024] Step S24, extract key features from the processed data to obtain voltage change rate, current change rate, temperature gradient, stress-strain ratio and Gas concentration change rate, and fuse the associated modal features into a state space module.

[0025] Preferably, in step S3, the multi-physical field coupling modeling method is used to present the parameters of the power unit of the downhole robot in all aspects intuitively, including:

[0026] Constructing a stress-strain coupling model to monitor the deformation caused by the stress of the battery;

[0027] Constructing a model based on thermal-acoustic coupling to monitor the change of the internal temperature of the battery by simulating the change of the time delay, attenuation and dispersion of the ultrasonic signal caused by temperature; ​

[0028] An equivalent circuit model is constructed to predict the voltage and current of the battery, the internal resistance value of the battery is calculated based on the collected voltage and current, and a semi-empirical attenuation model is constructed based on the collected temperature data to predict the capacity of the battery;

[0029] A temperature-deformation coupling model is constructed to monitor the relationship between the internal temperature and deformation of the battery to monitor the deformation of the battery;

[0030] A thermal-electric coupling model is constructed to monitor the temperature distribution of the battery to determine the state of the internal current and voltage of the battery;

[0031] An electric-force coupling model is constructed to monitor the internal current and voltage of the battery in real time to prevent stress concentration caused by abnormal voltage and current;

[0032] A thermal-gas coupling model is constructed to ensure that thermal runaway is avoided in the early temperature rise stage of the battery, thereby preventing the generation of gas.

[0033] Preferably, in step S4, the future health state of the battery is predicted by a method of co-evolution of a multi-modal large model and a physical model, and the steps are as follows:

[0034] Step S41, feature extraction is performed from each coupling model, and the change rate, mean value, variance, maximum value, minimum value and the rising rate and change trend of the concentration of voltage, current, temperature, stress, strain are extracted respectively;

[0035] Step S42, the extracted features are mapped to the same semantic space for alignment, the features of different modalities are transformed through a fully connected layer, and the adjustment is optimized through a loss function;

[0036] Step S43, the aligned features of voltage, current, temperature, stress, strain and concentration are fused into a single joint representation using a series method to form a joint feature vector;

[0037] Step S44, after successful concatenation of the features, a linear transformation layer is used to reduce the dimension and create a unified fusion feature vector;

[0038] Step S45, a ReLU activation function is selected to construct a regression network and train it, and the fusion feature vector is input into the regression network composed of a cross-modal attention network and a diffusion model to output the predicted values of the state parameters of the battery;

[0039] Step S46, during the prediction process, the multi-modal large model calls the physical coupling model for real-time inference calculation, and uses the output of the physical coupling model as a constraint condition to continuously optimize its own prediction results.

[0040] Preferably, in step S5, the internal and external gas content of the downhole robot is monitored using a layered monitoring architecture; a gas sensor is installed inside the robot, inside the explosion-proof cavity and on the top of the robot, and a real-time concentration mapping model is used to monitor the concentration of gas in real time.

[0041] Preferably, in step S6, a multi-parameter comprehensive early warning method is used to intelligently warn the power unit of the downhole robot, to identify potential safety risks in advance and provide sufficient time to take measures to reduce losses caused by accidents, including the following steps:

[0042] Step S61, construct a hybrid neural network: use a time convolution network to process the time series data of voltage, current and stress; convert the temperature data into a three-dimensional point cloud format and use a graph neural network to process the three-dimensional temperature point cloud data; use a graph isomorphism network to process the strain data of the battery; use a graph neural network to process the spatial distribution data of gas concentration and gas concentration;

[0043] Step S62, splice the output of the physical coupling model, the output of the multi-modal large model and the real-time monitoring data, splice the output of the real-time mapping model and the features of the real-time monitoring data, form two comprehensive feature vectors, and process the comprehensive feature vectors through a fully connected layer;

[0044] Step S63, label the historical data, mark out three different levels of states: normal state, abnormal state and serious abnormal state, and train the hybrid neural network to optimize the model parameters;

[0045] Step S64, according to the risk indicators output by the hybrid neural network model, divide the warning results into different risk levels, and generate corresponding system prompts according to the risk levels;

[0046] Step S65, deploy the trained hybrid neural network model, comprehensively consider the battery temperature, stress, current, voltage, gas concentration and gas concentration, and when one or more parameters change abnormally or are about to exceed the safety threshold, the system issues a corresponding warning.

[0047] Preferably, in step S7, when the downhole robot issues a warning, the emergency disposal module dynamically adjusts the execution unit of the physical entity running state to ensure the safety of the power unit of the downhole robot, including:

[0048] Install a fire extinguishing device inside the explosion-proof cavity of the robot, and when a thermal runaway warning occurs, control the fire extinguishing device in the twin space to prevent the battery from entering the burning or even exploding stage;

[0049] When deformation warning occurs, the emergency treatment module can perform real-time load reduction or control internal pressure growth, and cut off the working circuit when exceeding the set threshold;

[0050] When stress warning occurs, the emergency treatment module adjusts the battery module output power to avoid stress aggravation;

[0051] When voltage and current warning occurs during charging, the emergency treatment module dynamically adjusts the charging and discharging current, reduces the working load power when voltage and current warning occurs during working, and timely cuts off the current when the situation is difficult to control;

[0052] When the battery capacity warning occurs, the emergency treatment module controls and reduces the battery discharge rate, limits the execution of high-power tasks, and automatically switches to a low-power consumption mode;

[0053] When temperature warning occurs, the emergency treatment module starts the intrinsically safe thermoelectric cooling system, efficiently absorbs the heat generated by the battery, and conducts it to the surrounding environment, thereby rapidly reducing the battery temperature;

[0054] When the gas concentration outside the robot exceeds the set threshold, the emergency treatment module immediately stops the movement of the robot;

[0055] When the gas concentration inside the robot exceeds the set threshold, the emergency treatment module cuts off the power supply of all electrical equipment;

[0056] When the gas concentration inside the explosion-proof compartment of the robot exceeds the set threshold, the emergency treatment module triggers the power-off locking mechanism of the explosion-proof compartment to cut off the power supply of all electrical equipment in the compartment.

[0057] Preferably, a fire extinguishing device is installed inside the explosion-proof compartment of the robot, and the fire extinguishing port of the fire extinguishing device is aligned with the battery, and high-concentration liquid nitrogen is used as the fire extinguishing medium in the fire extinguishing device.

[0058] The application also provides a digital twin monitoring system for a downhole robot power unit, which is composed of a battery, a multi-sensor module, a battery digital twin model, a data processing module, a physical modeling module, a multi-modal large model, a gas sensor, a warning module and an emergency treatment module;

[0059] Battery: a power element for powering the downhole robot, which is a downhole robot power unit and is located inside the explosion-proof compartment of the downhole robot;

[0060] Multi-sensor module: used for monitoring the internal physical state of the battery and collecting data therefrom;

[0061] Battery digital twin model: used for real-time mapping of the battery in the physical space in the twin space;

[0062] Data processing module: for processing the data collected by the sensor module;

[0063] Physical modeling module: for physical modeling of the processed data;

[0064] Multi-modal large model: for predicting the future health status of the battery;

[0065] Gas sensor: for monitoring the gas concentration inside, outside and in the explosion-proof cavity of the robot;

[0066] Early warning module: for analyzing the state of the battery, discovering potential risks in advance, and issuing timely warnings;

[0067] Emergency disposal module: for quick response and effective intervention measures, by controlling the work of various safety elements, to minimize accidents and losses.

[0068] Therefore, the present application proposes a downhole robot power unit digital twin monitoring method and system, which has the following beneficial effects:

[0069] (1) The data collected by the multi-sensor fusion can provide more abundant information, support more accurate fault diagnosis and early warning functions, and the ultrasonic acoustic elasticity non-destructive testing technology can monitor the physical state inside the battery pack core which is difficult to obtain by conventional sensors and physical observation methods.

[0070] (2) The wavelet transform method is used for denoising, which can retain low-frequency signals such as voltage trends, and when processing current signals, it can remove background noise while retaining current peaks and mutations, and when processing battery temperature data in a high-noise environment, it can dynamically adjust the threshold value through an adaptive threshold method to better remove noise.

[0071] (3) The method of coupling modeling of physical fields is adopted, considering the interaction between multiple physical processes such as electricity, heat and force, which can more truly reflect the dynamic behavior of the system under actual working conditions. Not only significantly improves the modeling accuracy, but also reveals the potential failure mechanism of the battery.

[0072] (4) The gas monitoring is carried out by adopting a hierarchical monitoring architecture, which can timely discover the gas problems inside the robot, the cavity layer monitoring controls the concentration change in the explosion-proof cavity, and the external layer monitoring masters the external environment, so as to discover hidden dangers in advance.

[0073] (5) The method of multi-modal large model and physical model cooperation is adopted, which can use the output of the physical coupling model as a constraint condition to continuously optimize its own prediction results, so as to realize high-precision prediction and analysis of complex systems and improve the accuracy and reliability of the model.

[0074] (6) The battery is prewarned by using a multi-parameter comprehensive prewarning method, the one-sidedness caused by single physical field analysis is avoided, the accuracy of prewarning is improved, and potential abnormal states and fault signs in the system can be captured in advance, so that early prewarning is realized.

[0075] (7) The battery is emergently treated through the emergency treatment module, the emergency plan can be quickly started, effective measures can be taken in the shortest time, and the loss of the underground robot and the mine is minimized.

[0076] (8) The fire extinguishing device is installed in the underground robot, which helps to quickly suppress combustion or explosion when the battery is in thermal runaway, prevents the spread of fire and the diffusion of harmful gas, and greatly improves the intrinsic safety of the robot in a closed and high-risk environment.

[0077] The technical solutions of the present application will be further described in detail below by means of the drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0078] Figure 1 is a flow chart of a digital twin monitoring method of a power unit of an underground robot according to the present application;

[0079] Figure 2 is a structural diagram of a digital twin monitoring system of a power unit of an underground robot according to the present application;

[0080] Figure 3 is a structural diagram of an explosion-proof partition cavity according to the present application;

[0081] Figure 4 is a multi-modal data processing flow chart according to the present application;

[0082] Figure 5 is a battery health state prediction flow chart according to the present application;

[0083] Figure 6 is an intelligent prewarning flow chart of a power unit of an underground robot according to the present application.

[0084] REFERENCE NUMERALS

[0085] 1, gas sensor; 2, fire extinguishing device; 3, explosion-proof partition cavity; 4, ultrasonic longitudinal wave sensor; 5, ultrasonic surface wave sensor; 6, laser projection device; 7, gas sensor; 8, CCD camera; 9, intrinsically safe thermoelectric cooling system; 10, battery; 11, voltage and current monitoring module. DETAILED DESCRIPTION

[0086] In order to make the technical solutions, advantages and objectives of the present application clearer, the technical solutions of the embodiments of the present application will be described clearly and completely below. The described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the described embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without any creative effort fall within the protection scope of the present application.

[0087] Unless otherwise defined, the technical terms or scientific terms used in the present application shall have the usual meanings understood by those of ordinary skill in the art to which the present application belongs.

[0088] Embodiment one

[0089] As shown in the flow chart of a downhole robot power unit digital twin monitoring method of the present application, the specific steps are as follows: Figure 1

[0090] S1, multi-modal data acquisition of stress, temperature, electrical, deformation and thermal runaway related parameters of the battery, the specific operation is as follows:

[0091] The downhole robot power unit is monitored and data is collected by using an ultrasonic surface receiving sensor, a voltage and current monitoring module 11, a method based on projected laser scanning and a gas sensor 7, including:

[0092] Stress data and temperature data acquisition: a method based on ultrasonic acoustic elasticity for battery pack internal stress monitoring is used to characterize the change characteristics of the complex stress field inside the battery, the ultrasonic sensor is composed of an ultrasonic longitudinal wave sensor 4 and an ultrasonic surface wave sensor 5, the ultrasonic longitudinal wave sensor 4 is symmetrically installed on both sides of the battery 10 to obtain the stress state inside the battery 10; two ultrasonic surface wave sensors 5 are installed on the upper surface of the battery 10 to obtain the surface tension state of the battery pack;

[0093] Electrical parameter acquisition: the voltage and current monitoring module 11 is used to collect the voltage and current inside the battery 10 to ensure that overcharging, overdischarging and even short circuit problems do not occur inside the battery 10; the internal resistance value of the battery 10 can be calculated according to the collected data, and the capacity of the battery 10 can be indirectly judged by combining the temperature data of the battery 10;

[0094] Deformation data acquisition: a method based on projected laser scanning is used, the laser projection device 6 and the CCD camera 8 are respectively installed on the explosion-proof partition cavity inner wall above the battery 10, the optical axis of the CCD camera 8 needs to be perpendicular to the reference plane, when the battery 10 deforms due to complex working conditions, the CCD camera 8 senses the deformation of the battery 10 by collecting the projected stripes of the battery surface deformation.

[0095] ​Thermal runaway monitoring: gas sensor 7 is used to predict and monitor the thermal runaway of battery 10. Gas sensor 7 is installed on the inner wall of the explosion-proof cavity. By monitoring the gas concentration generated by the thermal runaway inside battery 10, the early thermal runaway of battery 10 can be predicted.

[0096] S2, as shown, the collected multi-modal data is pre-processed, and the pre-processed data is transmitted into the digital twin space, and the steps are as follows: Figure 4

[0097] S21, the signal is decomposed into different frequency components by wavelet transform to remove noise, and then different modal data is aligned according to the time stamp;

[0098] S22, identify and eliminate outliers caused by measurement errors, and use multi-source data fusion method to fill in the data loss caused by complex communication environment underground;

[0099] S23, the minimum-maximum normalization method is used to normalize the voltage, current, temperature and gas concentration related data, the Z-score normalization method is used to normalize the data with approximately normal distribution of stress and strain, and the data is converted to a unified scale, and the normalized data is stored in a unified format;

[0100] S24, extract the key features of the processed data to get the voltage change rate, current change rate, temperature gradient, stress-strain ratio and gas concentration change rate, and fuse the associated modal features into a state space module.

[0101] S3, build a twin model of the battery in the twin space and physically model the incoming data to monitor the current battery state;

[0102] The method of multi-physical field coupling modeling is used to intuitively present the parameters of the power unit of the downhole robot in all aspects to cope with the influence of the complex and changeable situation underground, including:

[0103] Construct a stress-strain coupling model to monitor the deformation caused by the stress of battery 10 to prevent cell deformation and shell rupture;

[0104] Construct a model based on thermal-acoustic coupling to monitor the change of temperature inside battery 10 by simulating the change of time delay, attenuation and dispersion of ultrasonic signal caused by temperature;

[0105] Construct an equivalent circuit model to predict the voltage and current of battery 10 to ensure that the battery will not be overcharged, overdischarged and short-circuited, calculate the internal resistance value of battery 10 by collecting the voltage and current, and combine the collected temperature data to construct a semi-empirical attenuation model to estimate the capacity of battery 10 to optimize the management of battery 10;​​

[0106] Constructing a temperature-deformation coupling model to monitor the relationship between the temperature and deformation inside the battery 10 to monitor the deformation of the battery 10;

[0107] Constructing a thermal-electric coupling model to monitor the temperature distribution of the battery 10 to determine the state of the current and voltage inside the battery 10;

[0108] Constructing an electric-force coupling model to monitor the current and voltage inside the battery 10 in real time to prevent stress concentration caused by abnormal voltage and current;

[0109] Constructing a thermal-gas coupling model to monitor the early temperature causes of the battery 10 that lead to thermal runaway , to ensure enough time to take appropriate measures to prevent thermal runaway and even explosion of the power unit of the downhole robot.

[0110] S4, as shown in Figure 5 , the health status of the battery is deduced by the method of cooperation between the multi-modal large model and the physical model, and the future health status of the battery is predicted, and the steps are as follows:

[0111] S41, feature extraction is performed from each coupling model, and the change rate, mean, variance, maximum value, minimum value and concentration of voltage, current, temperature, stress, strain are extracted respectively, which are used for subsequent alignment processing;

[0112] S42, the extracted features are mapped to the same semantic space for alignment, and the features of different modalities are transformed through a fully connected layer, so that their dimensions and scales can be correctly matched to ensure compatibility with the next step of fusion, and the loss function is optimized and adjusted;

[0113] S43, using a series method, the aligned features of voltage, current, temperature, stress, strain and concentration are fused into a single joint representation to form a joint feature vector;

[0114] S44, after successful concatenation of features, a linear transformation layer is used to reduce the dimension and create a unified feature vector, so that the multi-modal large model can learn the best representation of the fused features.

[0115] S45, select ReLU activation function to construct regression network and train it, input the fused feature vector into the regression network composed of cross-modal attention network and diffusion model, and output the predicted value of each state parameter of the battery;

[0116] S46, during the prediction process, the multi-modal large model calls the physical coupling model for real-time inference calculation, and uses the output of the physical coupling model as a constraint condition to continuously optimize its own prediction results.

[0117] S5, a gas sensor is installed outside the downhole robot and in the explosion-proof cavity to monitor the gas concentration when the downhole robot is working, specifically:

[0118] The layered monitoring architecture is used to monitor the gas content inside and outside the downhole robot; a gas sensor is installed inside the robot, inside the explosion-proof cavity 3 and above the outside of the robot to monitor the gas concentration; the external gas sensor ensures that the robot will not work in an area where the gas concentration exceeds the safety value; the internal gas sensor of the robot is used to ensure the sealing of the internal space of the robot; the internal gas sensor 1 of the explosion-proof cavity is used to ensure that the gas concentration will not explode due to abnormal power supply of the battery 10 or electric sparks generated during work, and a concentration real-time mapping model is used to monitor the concentration of gas in real time.

[0119] S6, as shown in Figure 6 When the predicted value of the health state of the battery and the gas concentration are about to exceed the safety threshold, the warning system gives a warning prompt in the twin space, and the specific steps are as follows:

[0120] S61, construct a hybrid neural network: use a time convolution network to process the time series data of voltage, current and stress; convert the temperature data into a three-dimensional point cloud format, and use a graph neural network to process the three-dimensional temperature point cloud data; use a graph isomorphism network to process the strain data of the battery; use a graph neural network to process the spatial distribution data of gas concentration and gas concentration;

[0121] S62, splice the output of the physical coupling model, the output of the multi-modal large model and the real-time monitoring data, splice the output of the real-time mapping model and the features of the real-time monitoring data, form two comprehensive feature vectors, and process the comprehensive feature vectors through a fully connected layer;

[0122] S63, label the historical data, mark the three different levels of states of normal state, abnormal state and serious abnormal state, and train the hybrid neural network to optimize the model parameters;

[0123] S64, according to the risk indicators output by the hybrid neural network model, divide the warning results into different risk levels, and generate corresponding system prompts according to the risk levels;

[0124] S65, deploy the trained hybrid neural network model, and comprehensively consider the battery temperature, stress, current, voltage, gas concentration and gas concentration, when one or more parameters change abnormally or are about to exceed the safety threshold, the system issues a corresponding warning.

[0125] S7, when the downhole robot occurs early warning, through the emergency disposal module dynamically adjusts the execution unit of the physical entity running state, ensures the safety of the downhole robot power unit, specifically includes:

[0126] The fire extinguishing device 2 is installed inside the robot explosion-proof partition cavity 3. When a thermal runaway early warning occurs, the battery is prevented from entering the burning or even explosion stage by controlling the fire extinguishing device 2 in the twin space; wherein the fire extinguishing device 2 uses high-concentration liquid nitrogen as fire extinguishing medium. When a thermal runaway early warning occurs, the low temperature of high-concentration liquid nitrogen can quickly reduce the surface temperature of the battery 10, block the thermal runaway chain, and nitrogen evaporation can form a protective layer on the surface of the battery 10 to isolate oxygen and inhibit the combustion reaction; through the dual effects of rapid cooling and oxygen deprivation, the early thermal runaway of the battery 10 is prevented and the battery 10 is prevented from entering the burning or even explosion stage.

[0127] When deformation early warning occurs, the emergency disposal module can perform real-time load reduction or control internal pressure growth, and cut off the working circuit when exceeding the set threshold;

[0128] When stress early warning occurs, the emergency disposal module adjusts the battery module output power to avoid stress aggravation;

[0129] When charging, when voltage and current early warning occurs, the emergency disposal module dynamically adjusts the charge and discharge current, when voltage and current early warning occurs during work, reduces the working load power, and when the situation is difficult to control, cuts off the current in time;

[0130] When the battery 10 capacity early warning occurs, the emergency disposal module controls and reduces the battery 10 discharge rate, limits the execution of high-power tasks, and automatically switches to low-power mode;

[0131] When temperature early warning occurs, the emergency disposal module starts the intrinsically safe thermoelectric cooling system 9, efficiently absorbs the heat generated by the battery 10, and conducts it to the surrounding environment, thereby rapidly reducing the temperature of the battery 10;

[0132] When the gas concentration outside the robot exceeds the set threshold, the emergency disposal module immediately stops the movement of the robot;

[0133] When the gas concentration inside the robot exceeds the set threshold, the emergency disposal module cuts off the power supply of all electrical equipment;

[0134] When the gas concentration inside the robot explosion-proof partition cavity 3 exceeds the set threshold, the emergency disposal module triggers the power-off locking mechanism of the explosion-proof partition cavity 3, cuts off the power supply of all electrical equipment in the explosion-proof partition cavity 3, and prevents electric sparks from causing gas explosion.

[0135] Example two

[0136] As Figure 2As shown, the present invention also provides a digital twin monitoring system for a downhole robot power unit, which is composed of a battery, a multi-sensor module, a battery digital twin model, a data processing module, a physical modeling module, a multimodal large model, a gas sensor 1, an early warning module, and an emergency response module;

[0137] Battery: a power element used to power the underground robot, serving as the underground robot power unit, located inside the underground robot explosion-proof compartment 3;

[0138] The internal layout of the underground robot explosion-proof compartment 3 is as follows Figure 3 As shown, it consists of a gas sensor 1, a fire extinguishing device 2, an ultrasonic longitudinal wave sensor 4, an ultrasonic surface wave sensor 5, a laser projection device 6, a gas sensor 7, a CCD camera 8, an intrinsically safe thermoelectric cooling system 9, a battery 10 and a voltage and current monitoring module 11.

[0139] Multi-sensor module: used to monitor the internal physical state of the battery 10 and collect data thereof;

[0140] Battery digital twin model: used to map the battery 10 in the physical space in real time in the twin space;

[0141] Data processing module: used to process the data collected by the sensor module;

[0142] Physical modeling module: used to perform physical modeling on the processed data;

[0143] Multimodal large model: used to predict the future health status of the battery 10;

[0144] Gas sensor 1: used to monitor the gas concentration inside and outside the robot and inside the explosion-proof compartment 3;

[0145] Early warning module: used to analyze the status of the battery 10, detect potential risks in advance, and issue an alarm in time;

[0146] Emergency response module: used for rapid response and effective intervention measures, minimizing the occurrence of accidents and losses by controlling the operation of various safety components.

[0147] It is worth noting that the contents not elaborated in detail in the present invention are all prior art and are well known to those skilled in the art.

[0148] Therefore, this paper provides a digital twin monitoring method and system for the power unit of an underground robot. This system uses multiple sensors to collect battery data, constructs a multi-physics field coupling model, and integrates multimodal large-scale model analysis technology to achieve efficient and accurate monitoring and trend prediction of the power unit status. Furthermore, gas sensors are used to ensure a safe operating environment, and combined with early warning and emergency response modules, the system ensures the stable operation of the underground robot under complex working conditions.

[0149] It should be pointed out finally that the above embodiments are only used to illustrate the technical solutions of the present application but not to limit it, and although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can still be modified or replaced equivalently, and these modifications or equivalent replacements should not make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present application.

Claims

1. A digital twin monitoring method for an underground robot power unit, characterized in that: Here are the steps: Step S1, performing multimodal data collection on battery stress, temperature, electricity, deformation, and thermal runaway related parameters; Step S2: preprocess the collected multimodal data and transfer the preprocessed data into the digital twin space; Step S3: Build a twin model of the battery in the twin space and perform physical modeling on the incoming data to monitor the current battery status; Step S4: deduce the battery's health status through a multimodal large model and predict the battery's future health status; Step S5: Installing gas sensors on the outside of the underground robot and inside and outside the explosion-proof compartment to monitor the gas concentration when the underground robot is working; Step S6: When the predicted health status value of the battery and the gas concentration are about to exceed the safety threshold, an early warning prompt is issued in the twin space through the early warning system; Step S7: After the warning occurs, the safety of the underground robot working underground is ensured by controlling the operation of various safety elements; In step S3, the multi-physics field coupling modeling method is used to intuitively present the power unit parameters of the downhole robot in all aspects, including: Construct a stress-strain coupling model to monitor the deformation caused by battery stress; A thermoacoustic coupling-based model was constructed to monitor the internal temperature of the battery by simulating the changes in ultrasonic signal delay, attenuation, and dispersion caused by temperature. Construct an equivalent circuit model to predict the battery voltage and current. Calculate the battery's internal resistance using the collected voltage and current. Combined with the collected temperature data, construct a semi-empirical decay model to estimate the battery's capacity. Construct a temperature-deformation coupling model to monitor the relationship between the internal temperature and deformation of the battery to monitor the deformation of the battery; Build a thermal-electric coupling model to monitor the temperature distribution of the battery to determine the current and voltage status inside the battery; Construct an electric-mechanical coupling model to monitor the internal current and voltage of the battery in real time to prevent stress concentration caused by abnormal voltage and current; Constructing a thermal-gas coupling model to ensure that thermal runaway is avoided in the early temperature rise stage of the battery, thereby preventing Gas production; In step S4, the future health status of the battery is predicted by the collaborative method of the multimodal large model and the physical model. The steps are as follows: Step S41: Extract features from each coupling model, and extract the rate of change, mean, variance, maximum, minimum and The rate of increase and trend of concentration; Step S42: Map the extracted features to the same semantic space for alignment, transform the features of different modalities through a fully connected layer, and optimize and adjust them through a loss function; Step S43: Use the series method to connect voltage, current, temperature, stress, strain and The aligned features of the concentrations are fused into a single joint representation, forming a joint feature vector; Step S44: After the features are successfully concatenated, a linear transformation layer is used to reduce the dimension and create a unified fused feature vector; Step S45: Select the ReLU activation function to construct a regression network and train it, input the fused feature vector into the regression network composed of the cross-modal attention network and the diffusion model, and output the predicted values ​​of each battery state parameter; Step S46: During the prediction process, the multimodal large model calls the physical coupling model to perform real-time reasoning and calculation, using the output of the physical coupling model as a constraint to continuously optimize its own prediction results; In step S6, a multi-parameter comprehensive early warning method is used to provide intelligent early warning for the power unit of the underground robot, identify potential safety risks in advance, and take measures to reduce losses caused by accidents, including the following steps: Step S61, constructing a hybrid neural network: using a time convolutional network to process the time series data of voltage, current, and stress; converting the temperature data into a 3D point cloud format, and using a graph neural network to process the 3D temperature point cloud data; using a graph isomorphism network to process the strain data of the battery; using a graph neural network to process Gas concentration and spatial distribution data of gas concentration; Step S62: Concatenate the output of the physical coupling model, the output of the multimodal large model, and the real-time monitoring data, and concatenate the output of the real-time mapping model and the features of the real-time monitoring data to form two comprehensive feature vectors, which are then processed through a fully connected layer. Step S63: annotate the historical data to mark three different levels of status: normal state, abnormal state, and severe abnormal state, and train the hybrid neural network to optimize the model parameters; Step S64: Classify the warning results into different risk levels according to the risk indicators output by the hybrid neural network model, and generate corresponding system prompts according to the risk levels; Step S65: deploy the trained hybrid neural network model, taking into account the battery temperature, stress, current, voltage, Gas concentration and methane concentration: when one or more of these parameters show abnormal change trends or are about to exceed the safety threshold, the system will issue a corresponding warning.

2. A digital twin monitoring method for a downhole robot power unit according to claim 1, characterized in that: In step S1, ultrasonic sensors, voltage and current detection modules, a method based on projection laser scanning, and gas sensors are used to monitor and collect data from the downhole robot power unit, including: Stress and temperature data acquisition: Ultrasonic sensors are used to monitor stress within the battery pack based on ultrasonic acoustic elasticity to characterize the changing characteristics of the complex stress field inside the battery. Electrical parameter acquisition: The voltage and current detection module is used to collect the internal voltage and current of the battery. The internal resistance of the battery is calculated based on the collected data, and the battery capacity is indirectly determined by combining it with the battery temperature data. Deformation data acquisition: A projection laser scanning method is used to collect projection stripes of battery surface deformation to sense battery deformation; Thermal runaway monitoring: Use gas sensors to predict and monitor the thermal runaway of batteries.

3. A digital twin monitoring method for a downhole robot power unit according to claim 1, characterized in that: In step S2, the collected multimodal data is preprocessed as follows: Step S21: decompose the signal into components of different frequencies by wavelet transform to remove noise, and then align different modal data according to timestamps; Step S22: Identify and eliminate outliers caused by measurement errors, and use a multi-source data fusion method to fill in the data loss caused by the complex communication environment underground; Step S23: Use the minimum-maximum normalization method to normalize the voltage, current, temperature and Normalize the gas concentration related data. Use the Z-score normalization method to normalize the stress and strain data with an approximately normal distribution, convert the data to a unified scale, and store the normalized data in a unified format. Step S24: extract key features from the processed data to obtain voltage change rate, current change rate, temperature gradient, stress-strain ratio and The gas concentration change rate is calculated and the associated modal features are fused into the state space modulus.

4. A digital twin monitoring method for a downhole robot power unit according to claim 1, characterized in that: In step S5, a layered monitoring architecture is used to monitor the gas content inside and outside the downhole robot; a gas sensor is installed inside the robot, inside the explosion-proof compartment, and above the robot, and a concentration real-time mapping model is used to monitor the gas concentration in real time.

5. The digital twin monitoring method for a downhole robot power unit according to claim 1, characterized in that: In step S7, when the underground robot issues an early warning, the emergency response module dynamically adjusts the execution unit of the physical entity's operating state to ensure the safety of the underground robot's power unit, specifically including: A fire extinguishing device is installed inside the robot's explosion-proof compartment. When a thermal runaway warning occurs, the fire extinguishing device is controlled in the twin space to prevent the battery from burning or even exploding. When a deformation warning occurs, the emergency response module can reduce the load or control the internal pressure increase in real time, and cut off the working line when the set threshold is exceeded; When a stress warning occurs, the emergency response module adjusts the battery module output power to avoid aggravation of stress; During charging, when the voltage and current have an early warning, the emergency response module dynamically adjusts the charge and discharge current. When the voltage and current have an early warning during operation, the workload power is reduced. When the situation is difficult to control, the current is cut off in time. When the battery capacity is low, the emergency response module controls and reduces the battery discharge rate, limits the execution of high-power tasks, and automatically switches to low-power mode. When a temperature warning occurs, the emergency response module activates the intrinsically safe thermoelectric cooling system to efficiently absorb the heat generated by the battery and transfer it to the surrounding environment, thereby quickly reducing the battery temperature; When the gas concentration outside the robot exceeds the set threshold, the emergency response module immediately stops the robot's movement; When the gas concentration inside the robot exceeds the set threshold, the emergency response module cuts off the power supply to all electrical equipment; When the gas concentration inside the robot's explosion-proof compartment exceeds the set threshold, the emergency response module triggers the power-off locking mechanism of the explosion-proof compartment, cutting off the power supply to all electrical equipment in the compartment.

6. A digital twin monitoring method for a downhole robot power unit according to claim 5, characterized in that: A fire extinguishing device is installed inside the explosion-proof compartment of the robot, ensuring that the fire extinguishing port of the fire extinguishing device is aimed at the battery, and high-concentration liquid nitrogen is used as the fire extinguishing medium in the fire extinguishing device.

7. A digital twin monitoring system for an underground robot power unit, characterized by: It consists of a battery, a multi-sensor module, a battery digital twin model, a data processing module, a physical modeling module, a multimodal large model, a gas sensor, an early warning module, and an emergency response module. Battery: A power element used to power the underground robot. It is the power unit of the underground robot and is located inside the explosion-proof compartment of the underground robot. Multi-sensor module: used to monitor the internal physical state of the battery and collect data; Battery digital twin model: used to map the battery in the physical space in real time in the twin space; Data processing module: used to process the data collected by the sensor module; Physical modeling module: used to perform physical modeling on the processed data; Multimodal large model: used to predict the future health status of the battery; Gas sensor: used to monitor the gas concentration inside and outside the robot and inside the explosion-proof compartment; Early warning module: used to analyze the battery status, detect potential risks in advance, and issue timely alarms; Emergency response module: used for rapid response and effective intervention measures, minimizing the occurrence of accidents and losses by controlling the operation of various safety components.

Images