Thermal imaging monitoring system based on digital twinning

Abstract

The present application relates to the technical field of computer-aided monitoring and simulation, in particular to a kind of thermal imaging monitoring system based on digital twinning, including perception and edge computing module, for obtaining temperature matrix from thermal imaging sensor, and temperature matrix is calculated to obtain regional level thermal instability index;It is also used to determine the risk level corresponding to the regional level thermal instability index, and generate thermal anomaly feature vector when the risk level is greater than the preset normal level;Twinning platform rapid response module is used to receive thermal anomaly feature vector, and locate virtual assets in digital twinning scene according to thermal anomaly feature vector, then according to the risk level in thermal anomaly feature vector, execute hierarchical response decision to virtual assets, the response mode of the present application surpasses the traditional lagging alarm, and changes into a kind of autonomous safety guarantee ability with foresight and immediacy.

Description

Technical Field

[0001] This invention relates to the field of computer-aided monitoring and simulation technology, specifically to a thermal imaging monitoring system based on digital twins. Background Technology

[0002] Digital twin technology monitors and optimizes the entire lifecycle of the physical world by constructing virtual images of physical entities. In the field of security monitoring, digital twin systems combined with thermal imaging sensors are used for fire early warning. Existing technical solutions generally follow a perception-transmission-analysis-decision process, transmitting thermal imaging video streams to a central server for analysis. This process suffers from data transmission redundancy, central computing bottlenecks, and decision-making logic lags, resulting in a system response arc time from anomaly detection to triggering a response that is typically on the order of seconds (approximately 10 seconds). 3 For sudden emergencies such as electrical fires, this delay can cause delays that prevent optimal intervention. Therefore, there is an urgent need in the field for a method that can compress response time to the millisecond level (approximately 10 ms). 2 The technology uses ms to achieve near-biological-instinctive heat-reflective safety protection. Summary of the Invention

[0003] The purpose of this invention is to provide a thermal imaging monitoring system based on digital twins, which solves the problems existing in the background technology.

[0004] To address the aforementioned technical problems, this invention provides a thermal imaging monitoring system based on digital twins, comprising:

[0005] The perception and edge computing module is used to acquire the temperature matrix from the thermal imaging sensor and calculate the temperature matrix to obtain the regional thermal instability index; it is also used to determine the risk level corresponding to the regional thermal instability index and generate a thermal anomaly feature vector when the risk level is greater than the preset normal level.

[0006] The rapid response module of the digital twin platform is used to receive thermal anomaly feature vectors, locate virtual assets in the digital twin scenario based on thermal anomaly feature vectors, and then perform graded response decisions on virtual assets based on the risk level in the thermal anomaly feature vectors.

[0007] Preferably, the process by which the sensing and edge computing module calculates the regional-level thermal instability index is as follows:

[0008] For each pixel in the temperature matrix, calculate the pixel-level thermal instability index; and for a preset region of interest, take the maximum value of the pixel-level thermal instability index of all pixels in the region to determine the region-level thermal instability index.

[0009] Preferably, the pixel-level thermal instability index is a comprehensive risk index obtained by weighted linear combination of the following physical quantities:

[0010] Absolute temperature derived from the temperature matrix;

[0011] The rate of change of temperature over time was calculated using the forward difference method on historical temperatures;

[0012] And the temperature spatial gradient magnitude obtained by calculating the temperature matrix of a single frame using image processing gradient operators.

[0013] Preferably, the risk level determination process is as follows:

[0014] Compare the regional thermal instability index with preset attention thresholds, warning thresholds, and critical thresholds;

[0015] If the regional thermal instability index is greater than the critical threshold, the risk level is determined to be critical.

[0016] If the regional thermal instability index is less than or equal to the critical threshold and greater than the warning threshold, the risk level is determined to be warning.

[0017] If the regional thermal instability index is less than or equal to the warning threshold and greater than the attention threshold, the risk level is determined to be attention.

[0018] If the regional thermal instability index is less than or equal to the attention threshold, the risk level is determined to be normal.

[0019] Preferably, the process of generating the thermal anomaly feature vector is as follows:

[0020] Obtain the sensor ID, anomaly coordinates, and event timestamp for addressing in the digital twin scenario; combine the risk level, regional thermal instability index, absolute temperature at the time of alarm triggering, temperature time change rate, and temperature spatial gradient magnitude.

[0021] To generate thermal anomaly feature vectors.

[0022] Preferably, the hierarchical response decision includes:

[0023] If the risk level is "attention", then the virtual assets will be highlighted in the twin model and the data refresh rate will be increased.

[0024] If the risk level is warning, then execute the response at the attention level and push alarm information to the operation and maintenance platform;

[0025] If the risk level is critical, the autonomous safety procedure will be triggered.

[0026] Preferably, the execution process of the autonomous safety procedure is as follows:

[0027] Switch the virtual asset to a preset emergency shutdown or electrical isolation state; invoke the lightweight simulation model, using the absolute temperature, temperature-time change rate, and temperature spatial gradient magnitude in the thermal anomaly feature vector as initial conditions, and deduce the anomaly development trend; based on the deduction results, issue control commands to the actuators in the physical world.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] (1) The response speed has been improved by orders of magnitude. By setting up a sensing and edge computing module, the temperature matrix is ​​analyzed in real time at the data source. Only when the risk level is determined to exceed the normal state is a byte-level thermal anomaly feature vector generated for transmission. This avoids the delay of massive data transmission and enables the twin platform's rapid response module to receive risk signals almost synchronously. Ultimately, the end-to-end time from the occurrence of physical anomalies to the execution of hierarchical response decisions in the digital twin world is compressed from seconds to milliseconds, winning a critical intervention window for defending against instantaneous emergencies such as electrical fires, and improving the safety level.

[0030] (2) The intelligence and precision of decision-making have reached a new level. The introduced thermal instability index model transforms isolated physical quantities into a dynamic and multi-dimensional risk insight. The model not only considers the current value of absolute temperature, but also incorporates the temperature time change rate reflecting the warming trend and the temperature spatial gradient modulus representing the degree of hot spot concentration. This shift from state monitoring to trend prediction enables the system to identify potential risks earlier. The execution of autonomous safety procedures is achieved by calling a lightweight simulation model and using real-time physical parameters in the thermal anomaly feature vector as initial conditions for deduction. This ensures that the control commands finally issued to the physical actuator are optimized for the current specific anomaly scenario, thus making the decision both fast and accurate.

[0031] (3) The system resource efficiency and scalability have been greatly optimized. The continuous transmission of the original video stream has been transformed into a lightweight hot anomaly feature vector that is only transmitted in case of anomalies. This has greatly saved network bandwidth and reduced the dependence on high-specification network infrastructure. The computing load has been distributed from a single central server to distributed sensing and edge computing modules, which not only reduces the processing pressure on the central platform, but also makes the entire monitoring system more scalable and can smoothly connect more monitoring points without creating performance bottlenecks.

[0032] (4) The model has excellent interpretability and environmental adaptability. Based on the statistical analysis and optimization algorithm of specific scenario datasets, the entire decision-making process is logically clear, traceable and verifiable. This transparency and parameterized design enables the scheme to flexibly adapt to the background noise of different industrial environments and the protection requirements of different safety levels by adjusting the weights and thresholds. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a logic block diagram of the system of the present invention;

[0035] Figure 2 This is a logic block diagram for calculating the regional thermal instability index in this invention;

[0036] Figure 3 This is a logical block diagram of the comprehensive risk indicators of the present invention;

[0037] Figure 4 This is a logic block diagram for determining the risk level of this invention. Detailed Implementation

[0038] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0039] Example 1

[0040] Please see Figure 1 This invention provides a thermal imaging monitoring system based on digital twins, comprising:

[0041] The perception and edge computing module is used to acquire the temperature matrix from the thermal imaging sensor and calculate the temperature matrix to obtain the regional thermal instability index; it is also used to determine the risk level corresponding to the regional thermal instability index and generate a thermal anomaly feature vector when the risk level is greater than the preset normal level.

[0042] The rapid response module of the digital twin platform is used to receive thermal anomaly feature vectors, locate virtual assets in the digital twin scenario based on thermal anomaly feature vectors, and then perform graded response decisions on virtual assets based on the risk level in the thermal anomaly feature vectors.

[0043] In this embodiment, the digital twin-based thermal imaging monitoring system aims to overcome the inherent data transmission redundancy and central computing bottlenecks in the existing sensing-transmission-analysis-decision link. To achieve this, the system places core analytical capabilities at the sensing and edge computing modules at the monitoring site and establishes an efficient edge-central collaboration mechanism. The final technical effect is that the previously prevalent second-level safety response time is successfully compressed to the millisecond level. This order-of-magnitude speed improvement enables the system to intervene in instantaneous hazards such as electrical fires in near real-time. Its response mode surpasses the traditional delayed alarm and transforms into a proactive and immediate autonomous safety assurance capability. The overall architecture optimization of the system allows the sensing and edge computing modules to focus on real-time data processing and risk prediction, while the twin platform's rapid response module focuses on accurate and tiered handling after receiving lightweight instructions. The two work together to significantly improve the system's resource efficiency and the level of intelligent decision-making.

[0044] Example 2

[0045] Please see Figure 2 The process by which the perception and edge computing module calculates the regional-level thermal instability index is as follows:

[0046] For each pixel in the temperature matrix, calculate the pixel-level thermal instability index; and for a preset region of interest, take the maximum value of the pixel-level thermal instability index of all pixels in the region to determine the region-level thermal instability index.

[0047] Please see Figure 3 The pixel-level thermal instability index is a comprehensive risk index obtained by weighted linear combination of the following physical quantities:

[0048] Absolute temperature derived from the temperature matrix;

[0049] The rate of change of temperature over time was calculated using the forward difference method on historical temperatures;

[0050] And the temperature spatial gradient magnitude obtained by calculating the temperature matrix of a single frame using image processing gradient operators;

[0051] In this embodiment, the core of the risk assessment performed by the sensing and edge computing module is to abandon the traditional static temperature threshold judgment and instead adopt a more dynamic and predictive, customized thermal instability index. The technical motivation behind this design is that a single absolute temperature value cannot effectively distinguish between normal temperature rises and early signs of danger, while a comprehensive risk indicator can provide deeper insights.

[0052] Its underlying logic lies in the thermal instability index. The model is based on the first law of thermodynamics and existing theories in fire science regarding the critical conditions for thermal runaway. This invention aims to construct a composite index that can predict the probability of an immediate disaster by quantifying several key physical dimensions that are strongly related to the thermal runaway process, namely the absolute value of temperature, the rate of change over time, and the steepness of spatial distribution, thereby realizing the transformation from passive response to active prediction.

[0053] Pixel-level thermal instability index The calculation is performed using the following formula:

[0054] ;

[0055] in, For time t, spatial location The dimensionless comprehensive risk score, as the final output of the formula, is directly related to the risk level.

[0056] The absolute temperature is derived directly from the temperature matrix acquired by the edge computing unit from the thermal imaging sensor at time t. The corresponding pixel value in;

[0057] This is the partial derivative of temperature with respect to time, i.e., the rate of change of temperature over time. In the discrete implementation, it is determined by the edge computing units based on the current frame and the previous time step. The temperature difference was approximately calculated using the forward difference method in numerical analysis. The data source was the internal caching and calculation of continuously acquired temperature data by the unit.

[0058] The spatial gradient of temperature, its modulus It reflects the degree of temperature difference between the hotspot and the surrounding environment. This parameter is calculated by the edge computing unit using standard image processing gradient operators, such as the Sobel operator, on the temperature matrix of a single frame.

[0059] These are dimensionless weighting coefficients, and the sum of the three is always 1. Their specific values ​​are obtained by offline training on historical or simulation datasets containing normal and multiple fault scenarios, with the optimization goal of maximizing the early detection rate and minimizing the false alarm rate, and by iteratively solving using algorithms such as grid search.

[0060] Let be the maximum-minimum normalization function, and its mathematical expression is: This function maps input parameters with different physical dimensions, such as temperature (K), rate of change of temperature over time (K / s), and magnitude of temperature spatial gradient (K / m), to a unified format. The dimensionless values ​​within the interval ensure the consistency of physical dimensions on both sides of the formula, where the parameters corresponding to each physical quantity are... and It is obtained by collecting baseline data for a sufficiently long period of time under normal operating conditions for a specific monitoring scenario;

[0061] In the actual deployment of the technical solution, the sensing and edge computing module first processes the temperature matrix. Calculate the pixel-level thermal instability index for each pixel within the range. Subsequently, for the k-th region of interest pre-defined by a person skilled in the art based on monitoring needs, such as for a specific device or a critical connection point. Its regional thermal instability index By execution The system uses a maximum value aggregation logic instead of an average value to determine the data. The direct technical effect of this is that it can ensure the monitoring system has the highest sensitivity to the most dangerous points in the monitoring area, effectively avoiding the possibility of missing key risk signals due to data smoothing or dilution. This series of calculations enables the system to make accurate judgments based on an interpretable and multi-dimensional quantifiable indicator, greatly improving the intelligence and accuracy of decision-making.

[0062] Example 3

[0063] Please see Figure 4 The process for determining the risk level is as follows:

[0064] Compare the regional thermal instability index with preset attention thresholds, warning thresholds, and critical thresholds;

[0065] If the regional thermal instability index is greater than the critical threshold, the risk level is determined to be critical.

[0066] If the regional thermal instability index is less than or equal to the critical threshold and greater than the warning threshold, the risk level is determined to be warning.

[0067] If the regional thermal instability index is less than or equal to the warning threshold and greater than the attention threshold, the risk level is determined to be attention.

[0068] If the regional thermal instability index is less than or equal to the attention threshold, the risk level is determined to be normal.

[0069] In this embodiment, the logic for determining the risk level is to use the continuous regional thermal instability index calculated in the aforementioned steps. Values ​​are mapped to discrete risk levels with explicit operational instructions. This process forms a crucial bridge from risk quantification to decision initiation; its judgment logic follows these rules: ;

[0070] when At that time, the level is judged as Level 3, which is critical;

[0071] when At that time, the level is determined to be level 2, which is a warning;

[0072] when At that time, the level is determined to be level 1, which means you should follow.

[0073] when At that time, the level is 0, which is normal;

[0074] A key implementation detail lies in the threshold. The setting of this threshold is not arbitrary, but has a scientifically derived calibration source. To ensure the feasibility of this threshold, technicians need to use labeled datasets containing multiple scenarios in the offline stage to analyze the receiver operation characteristic curve, i.e., the ROC curve, and scientifically set it according to the specific requirements of different security levels for sensitivity and specificity. The technical contribution of this hierarchical processing mechanism is that it makes the risk response strategy no longer singular and extensive, but achieves refined hierarchical management. It can mobilize resources of the corresponding level according to the severity of the risk, avoid unnecessary system interference and alarm fatigue of operation and maintenance personnel, and make decision-making more accurate and efficient.

[0075] Example 4

[0076] The process of generating thermal anomaly feature vectors is as follows:

[0077] Obtain the sensor ID, anomaly coordinates, and event timestamp for addressing in the digital twin scenario; combine the risk level, regional thermal instability index, absolute temperature at the time of alarm triggering, temperature time change rate, and temperature spatial gradient magnitude.

[0078] To generate thermal anomaly feature vectors;

[0079] In this embodiment, once the risk level is determined... Once determined to be greater than 0, the core task of the perception and edge computing module switches to generating a highly structured, information-condensed thermal anomaly feature vector. This vector is the fundamental carrier for achieving low-latency communication and accurate response in this invention; its structure is as follows:

[0080] ;

[0081] Each parameter within the vector has a clear origin and purpose: These are the sensor's unique identifier, the coordinates of the anomaly in the monitoring matrix, and the precise timestamp of the event. These originate from the edge computing unit's own configuration information and the system clock, and together they constitute the key index for spatial and temporal addressing in the digital twin platform.

[0082] The risk level is determined by the aforementioned logical relationship and serves as the direct basis for subsequent graded response decisions;

[0083] This set of parameters is a snapshot of the core physical quantities at the moment the alarm is triggered. They all come from the calculation process of the thermal instability index and will be used as the initial conditions for simulation in the subsequent twin refined response.

[0084] Based on this design, the direct technical consequence is that the data to be transmitted is drastically reduced from a MB-level thermal imaging video stream containing massive amounts of redundant information to a B-level feature vector of only tens of bytes. This order-of-magnitude data compression fundamentally eliminates network transmission bottlenecks, is one of the core prerequisites for achieving millisecond-level response, and greatly saves network bandwidth resources.

[0085] Example 5

[0086] Tiered response decision-making includes:

[0087] If the risk level is "attention", then the virtual assets will be highlighted in the twin model and the data refresh rate will be increased.

[0088] If the risk level is warning, then execute the response at the attention level and push alarm information to the operation and maintenance platform;

[0089] If the risk level is critical, the autonomous safety procedure will be triggered;

[0090] According to the aforementioned scheme, a thermal imaging monitoring system based on digital twins is characterized in that the execution process of the autonomous safety procedure is as follows:

[0091] Switch the virtual asset to a preset emergency shutdown or electrical isolation state; call the lightweight simulation model, using the absolute temperature, temperature-time change rate, and temperature spatial gradient magnitude in the thermal anomaly feature vector as initial conditions, and deduce the anomaly development trend; based on the deduction results, issue control commands to the actuators in the physical world.

[0092] In this embodiment, the twin platform's fast response module receives the thermal anomaly feature vector. Immediately afterwards, a risk level-related mechanism was activated. A strictly bound hierarchical response decision-making process; if For Level 1 attention, the virtual asset corresponding to the anomaly is only highlighted visually in the twin model, and its data refresh rate is increased for closer observation. This level does not trigger physical intervention. In the case of a Level 2 warning, in addition to executing a Level 1 response, the system will automatically push a regular alarm message to the operations and maintenance platform to notify relevant personnel to pay attention.

[0093] When the system faces the most urgent situation, that is In a Level 3 emergency, the highest priority autonomous safety procedure will be triggered. This procedure is a pre-programmed set of procedures designed for rapid intervention. The execution of this procedure demonstrates the core technology of this system's rapid closed-loop response: First, the system, according to... After locating the virtual asset using the addressing information, the system immediately switches its digital state to a preset safety state such as electrical isolation or emergency shutdown. Following this, the system invokes a lightweight, localized simulation model bound to the asset and decisively... The physical quantity carried in the snapshot T, , and As the initial condition for this simulation model, the development trend of the thermal anomaly over the next hundreds of milliseconds is rapidly deduced within milliseconds, and the effects of different intervention measures are simulated. Finally, based on the evaluation results of this micro-simulation, the system issues optimized control commands to the corresponding actuators in the physical world, such as smart circuit breakers or solenoid valves, to execute physical interventions. The entire process, from receiving vectors to issuing physical commands, achieves an end-to-end millisecond-level closed loop, constructing an intelligent safety protection system that is not only responsive but also makes accurate decisions and has autonomous closed-loop capabilities.

[0094] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A thermographic monitoring system based on digital twinning, characterized by, The method comprises the following steps: The perception and edge computing module is used to obtain a temperature matrix from a thermal imaging sensor and to calculate the temperature matrix to obtain a regional thermal instability index; The method is also used to determine a risk level corresponding to the regional thermal instability index, and to generate a thermal anomaly feature vector when the risk level is greater than a preset normal level; The twin platform rapid response module is used to receive the thermal anomaly feature vector, to locate a virtual asset in a digital twin scene according to the thermal anomaly feature vector, and to execute a hierarchical response decision on the virtual asset according to the risk level in the thermal anomaly feature vector; The process of calculating the regional thermal instability index by the perception and edge computing module is as follows: For each pixel in the temperature matrix, a pixel-level thermal instability index is calculated; and for a preset region of interest, the maximum value of the pixel-level thermal instability index of all pixels in the region is taken to determine the regional thermal instability index; The process of generating the thermal anomaly feature vector is as follows: The sensor ID, anomaly point coordinates and event timestamp used for addressing in the digital twin scene are obtained; the risk level, the regional thermal instability index, the absolute temperature at the time of triggering the alarm, the temperature time change rate and the temperature spatial gradient module length are combined; The thermal anomaly feature vector is generated; The pixel-level thermal instability index is a comprehensive risk indicator obtained by weighted linear combination of the following physical quantities: The absolute temperature derived from the temperature matrix; The temperature time change rate calculated by the forward difference method on the historical temperature; And the temperature spatial gradient module length calculated by the image processing gradient operator on a single frame of temperature matrix.

2. The thermal imaging monitoring system based on digital twinning of claim 1, wherein, The determination process of the risk level is as follows: The regional thermal instability index is compared with a preset attention threshold, a warning threshold and a critical threshold; If the regional thermal instability index is greater than the critical threshold, the risk level is determined to be critical; If the regional thermal instability index is less than or equal to the critical threshold and greater than the warning threshold, the risk level is determined to be warning; If the regional thermal instability index is less than or equal to the warning threshold and greater than the attention threshold, the risk level is determined to be attention; If the regional thermal instability index is less than or equal to the attention threshold, the risk level is determined to be normal.

3. The thermal imaging monitoring system based on digital twinning of claim 1, wherein, The hierarchical response decision includes: If the risk level is attention, the virtual asset is highlighted in the twin model and the data refresh rate is increased; If the risk level is warning, the response of the attention level is executed and alarm information is pushed to the operation and maintenance platform; If the risk level is critical, an autonomous safety procedure is triggered.

4. The thermal imaging monitoring system based on digital twinning of claim 3, wherein, The execution process of the autonomous safety procedure is as follows: The state of the virtual asset is switched to a preset emergency shutdown or electrical isolation state; a lightweight simulation model is called, and the absolute temperature, the temperature time change rate and the temperature spatial gradient module length in the thermal anomaly feature vector are used as initial conditions to deduce the abnormal development trend; according to the deduction result, control instructions are issued to the actuators in the physical world.

Images