An intelligent AI detection method based on infrared thermography and automobile diagnosis

Abstract

The application discloses an intelligent AI detection method based on infrared thermal images and automobile diagnosis, and the technical scheme of the application significantly improves the electric vehicle fire safety detection efficiency through multi-modal data fusion and intelligent analysis, the synchronous collection of double-band infrared and vehicle operation parameters constructs a multi-dimensional data basis, ensures the space-time consistency of thermal radiation characteristics and electrical state, secondly, the dynamic correlation model realizes the accurate triggering of abnormal current to thermal image enhancement, the partition gradient processing combined with dynamic threshold adjustment optimizes the detection sensitivity according to the thermal characteristics of different components, the three-dimensional thermal field reconstruction breaks through the limitation of surface detection, and can identify internal potential hot spots, finally, the closed-loop diagnosis is formed through the intelligent matching of the case library, so that the fault positioning accuracy reaches the component level, and the overall scheme solves the problem that the existing technology only relies on thermal infrared images for electric vehicle fire safety detection, and the contrast of the infrared thermal image is low and the detail resolution is poor.

Description

Technical Field

[0001] This invention relates to the field of intelligent detection technology, specifically to an intelligent AI detection method based on infrared thermography and vehicle diagnostics. Background Technology

[0002] As described in the published patent CN116823676B, the new energy vehicle industry is now entering a new stage of accelerated development, with a significant increase in the number of electric vehicles and a corresponding rise in the number of charging piles. To shorten charging time for users, charging piles generally adopt high-voltage, high-current operation, thus placing higher demands on the fire safety of electric vehicles. Infrared thermal imagers are a diagnostic technology that can instantly visualize and verify thermal information, possessing unique temperature measurement capabilities. When applied to electric vehicle fire safety detection, they can provide early warnings, detecting high-temperature points during the pre-ignition heat accumulation and smoldering stages, and sensing abnormal temperature changes caused by charging faults. Combined with smoke detectors, monitoring cameras with biometric recognition, and infrared curtain detectors, information can be collected to determine whether to activate emergency fireproof roller shutters, sprinkler heads, and sprinkler pumps, as well as provide voice announcements to ensure the charging safety of electric vehicles. However, infrared thermal images have drawbacks such as low contrast and poor detail resolution, affecting the accuracy of electric vehicle fire safety detection.

[0003] In summary, in the existing technology, relying solely on thermal infrared images for fire safety detection of electric vehicles has the problems of low contrast and poor ability to distinguish details. Summary of the Invention

[0004] To overcome the shortcomings mentioned above, this invention aims to provide a technical solution that can solve the aforementioned problems through an intelligent AI detection method based on infrared thermography and vehicle diagnostics.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] An intelligent AI detection method based on infrared thermal imaging and vehicle diagnostics includes the following steps:

[0007] S100: Simultaneously acquire dual-band infrared thermal imaging data and vehicle operating status parameters, the operating status parameters including at least battery pack voltage and current, motor temperature, and ambient temperature and humidity;

[0008] S200: Establish a dynamic correlation model between temperature and electrical parameters. When abnormal current fluctuations are detected, the infrared image contrast enhancement processing of the corresponding area is automatically triggered.

[0009] S300: Divide the vehicle thermal map into three detection areas: battery module area, high-voltage wiring harness area, and electronic control unit area, and perform gradient histogram equalization on each area.

[0010] S400: Based on real-time vehicle speed and ambient temperature, dynamically adjust the temperature anomaly detection threshold for each zone;

[0011] S500: When the abnormal features of infrared thermal images are spatiotemporally correlated with the abnormal electrical parameters, a three-dimensional heat conduction simulation is initiated to reconstruct the internal thermal field distribution and output the fault location and risk level assessment results.

[0012] S600: Combines historical fault case database for feature matching and optimizes subsequent matching priority.

[0013] As a further aspect of the present invention, step S100 specifically includes the following steps:

[0014] S110: Employs a dual-band infrared camera (medium wave and long wave) to simultaneously acquire vehicle thermal radiation images, with a sampling interval of 0.5-2 seconds between each frame.

[0015] S120: Real-time acquisition of voltage, current and individual cell temperature data of the battery management system (BMS) via the vehicle's CAN bus, and synchronous acquisition of IGBT module temperature and operating frequency parameters of the motor controller.

[0016] S130: Acquires real-time ambient temperature and humidity data of the vehicle's current location through onboard environmental sensors;

[0017] S140: Perform spatiotemporal alignment processing on infrared images and electrical parameter data.

[0018] As a further aspect of the present invention, step S200 specifically includes the following steps:

[0019] S210: Establish a current-temperature response baseline model and calculate the normal current fluctuation range of each battery module branch based on the vehicle's current SOC state and driving conditions.

[0020] S220: When the rate of change of a single branch current exceeds 150% of the baseline value for 5 seconds, the abnormal current event flag is activated.

[0021] S230: Based on the physical location index of the current abnormal branch, locate the area coordinates of the corresponding battery module in the infrared thermal map;

[0022] S240: Perform dynamic contrast enhancement processing and perform targeted enhancement processing on the target area;

[0023] S250: Generates an enhanced heatmap anomaly score. When the score exceeds a preset threshold, a three-level alarm protocol is triggered.

[0024] As a further aspect of the present invention, step S300 specifically includes the following steps:

[0025] S310: Based on feature matching between the vehicle structure CAD model and the heat map, the heat map is divided into:

[0026] Battery module area: Covers the outline of the battery pack casing and extends outward by 10% safety margin;

[0027] High-voltage harness area: A strip-shaped detection area with variable width is generated along the cable route;

[0028] Electrical control unit area: A rectangular detection frame with the geometric center of the motor controller housing as the reference;

[0029] S320: Perform differential gradient processing on each partition;

[0030] S330: Dynamically adjusts histogram equalization parameters based on real-time ambient temperature.

[0031] When the ambient temperature is greater than 35℃, the grayscale stretching range in high-temperature areas is limited.

[0032] When the ambient temperature is <0℃, enhance the contrast sensitivity in the low-temperature region;

[0033] S340: Generates a zone-enhanced heatmap and calculates the contrast enhancement coefficient for each region. If the coefficient does not meet the set standard, it triggers an image re-acquisition command.

[0034] As a further aspect of the present invention, step S400 specifically includes the following steps:

[0035] S410: Dynamically adjusts the temperature thresholds of each zone based on real-time vehicle speed, ambient temperature, and battery charging status, where:

[0036] The threshold value of the battery module area increases with vehicle speed and decreases with ambient temperature.

[0037] The threshold value of the high-voltage harness area is dynamically adjusted according to the charging current intensity.

[0038] The threshold values ​​of the electronic control unit area are inversely correlated with the ambient temperature.

[0039] S420: Set different threshold update frequencies according to the vehicle's driving or charging status;

[0040] S430: When the temperature of a certain partition exceeds the threshold continuously, perform additional verification steps corresponding to the partition type.

[0041] As a further aspect of the present invention, step S500 specifically includes the following steps:

[0042] S510: When an abnormal area in the thermal map and an abnormal electrical parameter are detected, acquire the component material parameters and adjacent sensor data for that area;

[0043] S520: Based on the surface temperature distribution in the anomalous region, a three-dimensional heat conduction model is established and the internal heat field propagation path is simulated;

[0044] S530: Generate a thermal risk diffusion trend map based on the simulation results and mark the locations of potential hot spots inside the room;

[0045] S540: Compare the simulation results with the historical fault feature database and output the fault type and risk level with the highest matching degree.

[0046] As a further aspect of the present invention, step S600 specifically includes the following steps:

[0047] S610: Extract current anomaly features, including temperature distribution patterns, electrical parameter change curves, and heat conduction trends;

[0048] S620: Perform similarity matching between the abnormal features and the historical fault case database, and calculate the matching score for each case;

[0049] S630: Select the top three cases with the highest matching scores, and generate a comprehensive risk level by assigning weights based on feature importance;

[0050] S640: Based on the fault location data of matched cases, generate a visual location probability map and mark high-incidence areas;

[0051] S650: Store the diagnosis result as a new case in the history database and optimize the matching priority for subsequent cases.

[0052] As a further aspect of the present invention, step S140 specifically includes the following steps:

[0053] S141: Add a unified timestamp to all data streams based on the vehicle clock source;

[0054] S142: Spatial registration of the infrared image coordinate system with the physical location of vehicle components;

[0055] S143: Establish a mapping matrix between changes in electrical parameters and changes in the heat map region.

[0056] As a further aspect of the present invention, step S240 specifically includes the following steps:

[0057] S241: Employs contrast adjustment processing to highlight the temperature difference details in the target area;

[0058] S242: Apply directional filtering along the extension direction of the battery module tabs to enhance the overheating characteristics of the connection area;

[0059] S243: Dynamically adjusts image enhancement intensity parameters based on ambient temperature.

[0060] As a further aspect of the present invention, step S320 specifically includes the following steps:

[0061] S321: The battery module area adopts lateral gradient reinforcement to highlight the thermal conduction characteristics of the connecting pieces between battery cells;

[0062] S322: Longitudinal filtering is implemented in the high-voltage harness area to enhance the detection of abnormal axial temperature distribution in cables;

[0063] S323: The electronic control unit area performs isotropic gradient calculations to capture local hot spots of components.

[0064] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0065] This invention significantly improves the fire safety detection efficiency of electric vehicles through multimodal data fusion and intelligent analysis. The synchronous acquisition of dual-band infrared and vehicle operating parameters constructs a multi-dimensional data foundation, ensuring the spatiotemporal consistency of thermal radiation characteristics and electrical status. Secondly, the dynamic correlation model enables the precise triggering of abnormal current for thermal image enhancement. Partition gradient processing combined with dynamic threshold adjustment optimizes the detection sensitivity for different component thermal characteristics. Three-dimensional thermal field reconstruction breaks through the limitations of surface detection and can identify potential internal hot spots. Finally, intelligent matching through a case library forms a closed-loop diagnosis, enabling fault location accuracy to reach the component level. The overall solution, through multi-level technical collaboration, solves the problems of low infrared thermal image contrast and poor detail resolution in existing technologies that rely solely on thermal infrared images for electric vehicle fire safety detection. Attached Figure Description

[0066] Figure 1 This is a flowchart of steps S100-S600 in this invention;

[0067] Figure 2 This is a flowchart of steps S110-S140 in this invention;

[0068] Figure 3 This is a flowchart of steps S210-S250 in this invention;

[0069] Figure 4 This is a flowchart of steps S310-S330 in this invention;

[0070] Figure 5 This is a flowchart of steps S410-S430 in this invention;

[0071] Figure 6 This is a flowchart of steps S510-S540 in this invention;

[0072] Figure 7 This is a flowchart of steps S610-S650 in this invention; Detailed Implementation

[0073] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0074] Please see Figure 1-7 A smart AI detection method based on infrared thermal imaging and vehicle diagnostics includes the following steps:

[0075] S100: Simultaneously acquire dual-band infrared thermal imaging data and vehicle operating status parameters, the operating status parameters including at least battery pack voltage and current, motor temperature, and ambient temperature and humidity;

[0076] S200: Establish a dynamic correlation model between temperature and electrical parameters. When abnormal current fluctuations are detected, the infrared image contrast enhancement processing of the corresponding area is automatically triggered.

[0077] S300: Divide the vehicle thermal map into three detection areas: battery module area, high-voltage wiring harness area, and electronic control unit area, and perform gradient histogram equalization on each area.

[0078] S400: Based on real-time vehicle speed and ambient temperature, dynamically adjust the temperature anomaly detection threshold for each zone;

[0079] S500: When the abnormal features of infrared thermal images are spatiotemporally correlated with the abnormal electrical parameters, a three-dimensional heat conduction simulation is initiated to reconstruct the internal thermal field distribution and output the fault location and risk level assessment results.

[0080] S600: Combines historical fault case database for feature matching and optimizes subsequent matching priority.

[0081] The intelligent AI detection method based on infrared thermal imaging and vehicle diagnostics in this invention achieves multi-level technical synergy in the field of electric vehicle fire safety by constructing a multi-dimensional collaborative detection system.

[0082] In the S100, during the data acquisition phase, the simultaneous acquisition of dual-band infrared imaging and vehicle operating parameters forms a complementary verification mechanism of multi-source heterogeneous data. Among them, mid-wave infrared enhances the capture of subtle temperature differences in metal components, while long-wave infrared improves the resolution of radiation characteristics of non-metallic materials. Combined with real-time operating parameters such as battery voltage and motor temperature, a complete data foundation for thermal-electrical-environment coupling analysis is constructed.

[0083] In S200, the establishment of the dynamic correlation model innovatively uses the transient characteristics of current as the trigger signal for image enhancement, realizing the mode transformation from passive full-domain scanning to active key monitoring. By prioritizing high-probability risk areas through directional enhancement algorithms, the efficiency of early hazard detection is significantly improved.

[0084] In S300, the regional gradient processing strategy fully considers the stacked thermal conduction characteristics of the battery module, the axial thermal diffusion law of the high-voltage wiring harness, and the concentrated heating characteristics of the electronic control unit. It adopts a differentiated spatial domain processing method to highlight the temperature gradient characteristics of each component that are most diagnostically valuable.

[0085] In the S400, the dynamic threshold adjustment mechanism introduces multi-dimensional variables such as vehicle speed and ambient temperature to build an adaptive compensation model, which effectively overcomes the adaptability defects of traditional fixed thresholds under complex working conditions such as high and low temperature alternation and rapid acceleration.

[0086] In the S500, the three-dimensional thermal field reconstruction technology breaks through the physical limitations of surface thermal imaging. By constructing an internal heat flow distribution model through reverse heat conduction calculation, it can accurately identify hidden thermal risks such as single cell failure inside the battery module and local breakdown of wire harness insulation layer.

[0087] In S600, the intelligent matching of historical case database forms a continuous evolution loop of diagnostic knowledge. While retaining typical fault characteristics, it continuously optimizes the feature weight allocation strategy through incremental learning, enabling the system to have the ability to reason and judge new fault modes.

[0088] This invention significantly improves the fire safety detection efficiency of electric vehicles through multimodal data fusion and intelligent analysis. The synchronous acquisition of dual-band infrared and vehicle operating parameters constructs a multi-dimensional data foundation, ensuring the spatiotemporal consistency of thermal radiation characteristics and electrical status. Secondly, the dynamic correlation model enables the precise triggering of abnormal current for thermal image enhancement. Partition gradient processing combined with dynamic threshold adjustment optimizes the detection sensitivity for different component thermal characteristics. Three-dimensional thermal field reconstruction breaks through the limitations of surface detection and can identify potential internal hot spots. Finally, intelligent matching through a case library forms a closed-loop diagnosis, enabling fault location accuracy to reach the component level. The overall solution, through multi-level technical collaboration, solves the problems of low infrared thermal image contrast and poor detail resolution in existing technologies that rely solely on thermal infrared images for electric vehicle fire safety detection.

[0089] In this embodiment of the invention, step S100 specifically includes the following steps:

[0090] S110: Employs a dual-band infrared camera (medium wave and long wave) to simultaneously acquire vehicle thermal radiation images, with a sampling interval of 0.5-2 seconds between each frame.

[0091] S120: Real-time acquisition of voltage, current and individual cell temperature data of the battery management system (BMS) via the vehicle's CAN bus, and synchronous acquisition of IGBT module temperature and operating frequency parameters of the motor controller.

[0092] S130: Acquires real-time ambient temperature and humidity data of the vehicle's current location through onboard environmental sensors;

[0093] S140: Perform spatiotemporal alignment processing on infrared images and electrical parameter data;

[0094] The technical solution of this invention lays a more reliable data foundation for the thermal safety testing of electric vehicles by constructing a precise acquisition system for multi-source heterogeneous data.

[0095] In the S110, the collaborative operation of dual-band infrared cameras enables complementary capture of the thermal radiation characteristics of metal components and non-metallic materials. Mid-wave infrared is used to sensitively capture the subtle temperature rise of metal structures such as battery connectors, while long-wave infrared is used to optimize the thermal anomaly analysis of non-metallic parts such as wire harness insulation layers. Combined with a dynamic sampling interval of 0.5-2 seconds, real-time performance is ensured while avoiding data redundancy.

[0096] In the S120, the innovative design of integrating BMS system data through the vehicle CAN bus deeply couples core electrical parameters such as battery cell voltage and current, IGBT module temperature with thermal imaging data, fully covering the status monitoring needs of key heat-generating nodes in the high-voltage system.

[0097] In S130, real-time acquisition of ambient temperature and humidity establishes an environmental compensation benchmark for thermodynamic analysis, effectively distinguishing between the actual temperature rise of components and environmental thermal radiation interference.

[0098] In S130, the spatiotemporal alignment processing eliminates the erroneous correlation of features caused by data acquisition delays and spatial misalignment in traditional methods by synchronizing the time of multi-source data and mapping spatial coordinates. This provides an accurate spatiotemporal reference for subsequent multimodal data analysis. This data acquisition system significantly improves the correlation effectiveness of thermal-electrical-environmental multidimensional data, enabling the system to leap from discrete parameter detection to comprehensive state assessment.

[0099] In this embodiment of the invention, step S200 specifically includes the following steps:

[0100] S210: Establish a current-temperature response baseline model and calculate the normal current fluctuation range of each battery module branch based on the vehicle's current SOC state and driving conditions.

[0101] S220: When a single branch current change rate is detected to exceed 150% of the baseline value and lasts for 5 seconds, the abnormal current event flag is activated.

[0102] S230: Based on the physical location index of the current abnormal branch, locate the area coordinates of the corresponding battery module in the infrared thermal map;

[0103] S240: Perform dynamic contrast enhancement processing and perform targeted enhancement processing on the target area;

[0104] S250: Generates an enhanced heatmap anomaly score. When the score exceeds a preset threshold, a three-level alarm protocol is triggered.

[0105] The technical solution of this invention realizes a multi-dimensional safety protection system in the field of electric vehicle battery safety monitoring by constructing an intelligent detection system that links current and thermal imaging.

[0106] In S210, the dynamic construction of the current-temperature response baseline model innovatively integrates the battery state of charge and real-time driving condition parameters to form a monitoring benchmark that adaptively adjusts according to the operating scenario, effectively solving the problem of misjudgment of traditional fixed thresholds under dynamic conditions such as acceleration and hill climbing.

[0107] In S220, the dual judgment mechanism for abnormal current events establishes reliable abnormal screening logic through the joint verification of the rate of change threshold and duration, ensuring sensitivity while avoiding false triggering caused by transient interference.

[0108] In S230, the topology mapping-based thermal mapping localization technology establishes the connection between the branch number of the electrical system and the spatial coordinates of thermal imaging, realizing a visual mapping from electrical anomaly signals to physical locations.

[0109] In S240, the targeted enhancement processing adopts an adaptive feature enhancement algorithm to perform local image optimization on key parts such as battery module connection interface and electrode solder joint, so that the thermal performance characteristics of micro-defects such as increased contact resistance can be highlighted.

[0110] In S250, the hierarchical alarm system establishes a progressive response mechanism from early warning to emergency intervention by quantitatively assessing the intensity of abnormal features in heat maps. It provides precise handling strategies for different risk levels. The entire solution, through the closed-loop collaboration of dynamic benchmark establishment, spatial correlation mapping, feature-oriented enhancement, and intelligent decision-making, not only improves the ability to detect hidden thermal faults, but also builds a hierarchical security protection system.

[0111] In this embodiment of the invention, step S300 specifically includes the following steps:

[0112] S310: Based on feature matching between the vehicle structure CAD model and the heat map, the heat map is divided into:

[0113] Battery module area: Covers the outline of the battery pack casing and extends outward by 10% safety margin;

[0114] High-voltage harness area: A strip-shaped detection area with variable width is generated along the cable route;

[0115] Electrical control unit area: A rectangular detection frame with the geometric center of the motor controller housing as the reference;

[0116] S320: Perform differential gradient processing on each partition;

[0117] S330: Dynamically adjusts histogram equalization parameters based on real-time ambient temperature.

[0118] When the ambient temperature is greater than 35℃, the grayscale stretching range in high-temperature areas is limited.

[0119] When the ambient temperature is <0℃, enhance the contrast sensitivity in the low-temperature region;

[0120] The technical solution of this invention achieves multi-dimensional performance optimization in the field of electric vehicle thermal imaging diagnosis by constructing an intelligent zoned detection system;

[0121] In the S310, the thermal mapping zoning strategy based on the vehicle structure CAD model innovatively combines engineering design with thermal radiation characteristics. The design of expanding the battery module area by 10% safety margin effectively captures the thermal diffusion phenomenon at the edge of the battery pack. The variable width strip division of the high-voltage harness area accurately adapts to the thermal conduction characteristics of cables of different diameters. The geometric positioning of the electronic control unit area ensures the integrity of temperature rise monitoring of core power devices.

[0122] In S320, the differentiated gradient processing implements targeted algorithm enhancement for the thermal distribution characteristics of each region. The lateral gradient enhancement of the battery module area can highlight the lateral heat flow abnormality of the electrode welding defect. The longitudinal filtering processing of the high voltage harness area sensitively captures the axial temperature gradient abrupt change caused by insulation layer damage. The isotropic processing of the electronic control unit area effectively identifies the radial thermal field characteristics of local overheating of the chip.

[0123] In the S330, the environmental temperature adaptive parameter adjustment mechanism breaks through the static mode of traditional image processing. In high-temperature environments, the grayscale control of the heat sink area avoids thermal noise interference, and the contrast enhancement in low-temperature environments significantly improves the visualization detection capability of low-temperature anomalies such as coolant leakage.

[0124] This solution systematically improves the reliability of thermal safety detection under complex operating conditions through the synergistic effect of precise spatial partitioning, targeted feature processing, and intelligent environmental compensation.

[0125] In this embodiment of the invention, step S400 specifically includes the following steps:

[0126] S410: Dynamically adjusts the temperature thresholds of each zone based on real-time vehicle speed, ambient temperature, and battery charging status, where:

[0127] The threshold value of the battery module area increases with vehicle speed and decreases with ambient temperature.

[0128] The threshold value of the high-voltage harness area is dynamically adjusted according to the charging current intensity.

[0129] The threshold values ​​of the electronic control unit area are inversely correlated with the ambient temperature.

[0130] S420: Set different threshold update frequencies according to the vehicle's driving or charging status;

[0131] S430: When the temperature of a certain partition continuously exceeds the threshold, execute the additional verification steps corresponding to the partition type;

[0132] The technical solution of this invention significantly improves the accuracy and reliability of thermal safety detection for electric vehicles by constructing a multi-dimensional adaptive temperature monitoring system.

[0133] In the S410, the dynamic threshold adjustment mechanism innovatively establishes a differentiated compensation strategy for different zones: the positive correlation between the threshold of the battery module area and the vehicle speed is used to match the reasonable temperature rise under high current discharge conditions, while the negative correlation with the ambient temperature effectively eliminates the risk of misjudgment of heat dissipation delay under high temperature conditions; the charging current correlation correction in the high voltage harness area accurately quantifies the allowable value of Joule thermal effect in fast charging scenarios; and the reverse coupling design between the threshold of the electronic control unit area and the ambient temperature cleverly balances the interactive influence of the heat dissipation system efficiency and the ambient heat load.

[0134] In the S420, the update frequency control of state awareness optimizes the allocation of system computing resources while ensuring real-time monitoring through differentiated management of driving and charging conditions.

[0135] In S430, a zoned customized verification strategy forms a hierarchical defense system: the temperature difference verification of adjacent cells in the battery module area can identify module imbalance faults; the axial gradient scanning in the high-voltage harness area can accurately locate insulation degradation points; and the heat dissipation performance correlation analysis in the electronic control unit area can effectively distinguish between actual overheating and cooling system failure. This solution achieves an intelligent upgrade of temperature anomaly detection under complex operating environments through multi-layer technology collaboration of dynamic benchmark construction, operating condition perception optimization, and defensive verification.

[0136] In this embodiment of the invention, step S500 specifically includes the following steps:

[0137] S510: When an abnormal area in the thermal map and an abnormal electrical parameter are detected, acquire the component material parameters and adjacent sensor data for that area;

[0138] S520: Based on the surface temperature distribution in the anomalous region, a three-dimensional heat conduction model is established and the internal heat field propagation path is simulated;

[0139] S530: Generate a thermal risk diffusion trend map based on the simulation results and mark the locations of potential hot spots inside the room;

[0140] S540: Compare the simulation results with the historical fault feature database and output the fault type and risk level with the highest matching degree.

[0141] The technical solution of this invention achieves a closed-loop technology from surface monitoring to essential analysis in the field of electric vehicle hidden danger detection by constructing a deep-sensing thermal safety diagnostic system.

[0142] In S510, the multi-source data fusion mechanism innovatively integrates physical property parameters such as material thermal conductivity and adjacent sensor readings, providing multi-dimensional data support for thermal field analysis and ensuring the completeness and engineering applicability of model input;

[0143] In S520, the three-dimensional thermal conduction modeling technology breaks through the surface monitoring limitations of traditional infrared detection. By reverse engineering the internal heat flow transfer path, it can accurately identify deep-seated hidden dangers such as thermal runaway of individual cells inside the battery module and latent breakdown of wire harness insulation layer.

[0144] In S530, the generation of the thermal risk diffusion trend map transforms the abstract thermodynamic simulation results into visualized spatial early warning information, and intuitively presents the evolution direction and potential outbreak point of the thermal threat through gradient coloring and contour line annotation.

[0145] In the S540, the intelligent case matching engine establishes a mapping network between thermodynamic features and fault types by mining deep association rules from the historical fault feature database. It can maintain high diagnostic confidence even in complex scenarios such as early short circuits and abnormal contact resistance. This solution forms a complete technical closed loop of surface anomaly detection, internal thermal field inference, and intelligent fault diagnosis through the dual-wheel synergy of physical modeling and data-driven approaches, which significantly improves the early warning capability of hidden thermal risks.

[0146] In this embodiment of the invention, step S600 specifically includes the following steps:

[0147] S610: Extract current anomaly features, including temperature distribution patterns, electrical parameter change curves, and heat conduction trends;

[0148] S620: Perform similarity matching between the abnormal features and the historical fault case database, and calculate the matching score for each case;

[0149] S630: Select the top three cases with the highest matching scores, and generate a comprehensive risk level by assigning weights based on feature importance;

[0150] S640: Based on the fault location data of matched cases, generate a visual location probability map and mark high-incidence areas;

[0151] S650: Store the diagnosis result as a new case in the history database and optimize the matching priority for subsequent cases.

[0152] The technical solution of this invention achieves the technical effect of continuous optimization from single detection in the field of electric vehicle thermal safety by constructing a self-evolving intelligent diagnostic system.

[0153] In S610, the innovative multi-dimensional feature extraction mechanism integrates temperature distribution, electrical parameter time-domain characteristics and heat conduction spatial patterns to form a three-dimensional abnormal feature description system, laying the foundation for multi-physics field coupling analysis for accurate diagnosis.

[0154] In S620, the intelligent case matching engine uses a dynamic weight allocation algorithm to mine deep correlations in historical data, capturing weak correlation signals of marginal cases while retaining typical fault characteristics.

[0155] In S630, the weighted decision-making strategy for the first three cases breaks through the limitations of traditional single matching and constructs a comprehensive risk prediction model with stronger anti-interference ability through multi-dimensional feature importance assessment.

[0156] In S640, the visualized location probability map transforms abstract diagnostic conclusions into spatial probability heat maps, and achieves an intuitive three-dimensional presentation of the fault location through heat value superposition and area coloring technology.

[0157] In S650, the dynamic update mechanism of the knowledge base forms a closed loop of continuous evolution of diagnostic capabilities, enabling the system to have progressive learning capabilities when dealing with new types of faults. This solution, through the full-chain collaboration of feature fusion analysis, intelligent case matching, comprehensive risk assessment, and self-optimization of diagnostic knowledge, not only improves the accuracy of fault identification but also builds an intelligent diagnostic ecosystem with autonomous evolution capabilities.

[0158] In this embodiment of the invention, step S140 specifically includes the following steps:

[0159] S141: Add a unified timestamp to all data streams based on the vehicle clock source;

[0160] S142: Spatial registration of the infrared image coordinate system with the physical location of vehicle components;

[0161] S143: Establish a mapping matrix between changes in electrical parameters and changes in heat map regions;

[0162] The technical solution of this invention achieves a unified and innovative spatiotemporal benchmark in the field of electric vehicle thermal safety detection by constructing a precise correlation framework for multi-source data.

[0163] In S141, the unified timestamp mechanism achieves millisecond-level time synchronization based on the vehicle clock source, completely eliminating the data time-domain misalignment caused by sampling frequency differences in traditional multi-sensor systems, and providing a reliable time-series benchmark for transient anomaly analysis.

[0164] In S142, the spatial registration of infrared images with physical locations uses a coordinate system transformation algorithm to map the pixels of the thermal image to the three-dimensional structural model of the vehicle, so that the abnormal thermal radiation areas can be accurately associated with specific components (such as battery cell numbers or wiring harness connector positions), and the positioning error is controlled within the engineering accuracy range of 3 centimeters.

[0165] In S143, the establishment of the electro-thermal mapping relationship matrix breaks through the limitations of single-modal data analysis. Through matrix operations, it reveals the quantitative correlation between current mutation and heat map gradient change. For example, it identifies the temperature rise rate characteristic of a specific area of ​​the heat map corresponding to every 10A increase in current of a certain branch as 0.8℃ / s. This technology system upgrades multimodal data from discrete acquisition to organic fusion, providing a spatiotemporally consistent and physically interpretable data foundation for subsequent intelligent diagnosis.

[0166] In this embodiment of the invention, step S240 specifically includes the following steps:

[0167] S241: Employs contrast adjustment processing to highlight the temperature difference details in the target area;

[0168] S242: Apply directional filtering along the extension direction of the battery module tabs to enhance the overheating characteristics of the connection area;

[0169] S243: Dynamically adjusts image enhancement intensity parameters based on ambient temperature;

[0170] The technical solution of this invention achieves a breakthrough improvement in the ability to resolve key area features in the field of electric vehicle thermal imaging detection by constructing a targeted intelligent enhancement system;

[0171] In S241, the temperature difference detail enhancement processing adopts an adaptive contrast adjustment algorithm, which highlights the micro temperature difference characteristics inside the battery module through nonlinear grayscale mapping, and improves the thermal performance intensity of micro-defects such as electrode tab soldering and coating peeling to the detectable range.

[0172] In S242, the directional filtering process innovatively deploys a spatial domain filtering kernel function along the battery tab axis. By allocating directional weights, it suppresses interference noise in the vertical direction, thereby improving the signal-to-noise ratio of the local overheating characteristics caused by the oxidation of the connecting piece by more than 3 times.

[0173] In S243, the dynamic parameter adjustment mechanism of ambient temperature sensing constructs a negative feedback control loop between image enhancement intensity and ambient thermal noise. Under high temperature conditions, it automatically suppresses over-enhanced artifacts in the heat sink area and intelligently improves the display contrast of condensation features of cooling pipes in low temperature environments. Through the synergy of three technologies—feature-oriented enhancement, noise-oriented suppression, and environmental adaptive adjustment—this solution can maintain the visualization detection sensitivity of key thermal hazards even under complex operating conditions.

[0174] In this embodiment of the invention, step S320 specifically includes the following steps:

[0175] S321: The battery module area adopts lateral gradient reinforcement to highlight the thermal conduction characteristics of the connecting pieces between battery cells;

[0176] S322: Longitudinal filtering is implemented in the high-voltage harness area to enhance the detection of abnormal axial temperature distribution in cables;

[0177] S323: The electronic control unit area performs isotropic gradient calculations to capture local hot spots of components;

[0178] The technical solution of this invention achieves a breakthrough improvement in multi-dimensional feature analysis capability in the field of electric vehicle thermal imaging detection by constructing a domain-specific directional enhancement system;

[0179] In S321, the horizontal gradient enhancement of the battery module area along the stacking direction of the battery cells strengthens the horizontal heat conduction characteristics, which improves the visualization detection sensitivity of horizontal heat flow anomalies such as electrode welding defects and connector corrosion to the micro-temperature difference level, effectively capturing contact resistance gradient faults that are difficult to identify by traditional methods.

[0180] In S322, the longitudinal filtering process of the high-voltage harness area deploys a spatial domain filtering kernel function along the cable axis. Through the synergistic effect of vertical noise suppression and axial feature enhancement, the signal-to-noise ratio of the longitudinal temperature gradient abrupt change characteristics caused by local breakdown of the harness insulation layer and joint oxidation is significantly improved.

[0181] In S323, the isotropic gradient calculation of the electronic control unit area adopts a multi-directional differential operator to synchronously detect the spatial change rate of the thermal field. It forms a 360-degree radial feature capture capability for point overheating defects such as IGBT module solder joint failure and capacitor bulging, breaking through the field-of-view limitation of traditional single-direction detection. Through the deep adaptation of spatial domain processing strategy and component thermodynamic characteristics, this scheme forms a systematic optimization in the detection efficiency of three typical thermal safety hazards: battery connection interface defects, wire harness axial deterioration, and local failure of electronic control components.

[0182] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. An intelligent AI detection method based on infrared thermography and car diagnostics, characterized in that, The method comprises the following steps: S100: synchronously collecting double-band infrared thermal imaging data and vehicle operating state parameters, wherein the operating state parameters at least include battery pack voltage and current, motor temperature, ambient temperature and humidity; S200: establishing a temperature-electric parameter dynamic correlation model, when detecting abnormal current fluctuation, automatically triggering infrared image contrast enhancement processing of the corresponding area; S300: dividing the vehicle thermal map into three detection areas of battery module area, high-voltage cable area and electronic control unit area, and performing gradient histogram equalization processing on each subarea respectively; S400: dynamically adjusting the temperature abnormality judgment threshold of each subarea based on real-time vehicle speed and ambient temperature; S500: when the infrared thermal map abnormal feature and the electric parameter abnormality have spatiotemporal correlation, starting three-dimensional heat conduction simulation to reconstruct the internal thermal field distribution, and outputting fault positioning and risk level evaluation results; S600: combining historical fault case library for feature matching, and optimizing subsequent matching priority. 2.The intelligent AI detection method based on infrared thermography and vehicle diagnosis of claim 1, wherein, The step S100 specifically comprises the following steps: S110: synchronously collecting vehicle thermal radiation images by using a middle wave and a long wave double-band infrared camera, and the sampling interval of each image is 0.5-2 seconds; S120: acquiring voltage and current data of a battery management system (BMS) and single battery temperature data through a vehicle CAN bus in real time, and synchronously collecting IGBT module temperature and working frequency parameters of a motor controller; S130: acquiring real-time ambient temperature and humidity data of the current position of the vehicle through a vehicle-mounted environment sensor; S140: performing spatiotemporal alignment processing on the infrared image and the electric parameter data. 3.The intelligent AI detection method based on infrared thermography and automotive diagnostics of claim 2, wherein, The step S200 specifically comprises the following steps: S210: establishing a current-temperature response baseline model, calculating the normal current fluctuation range of each battery module branch based on the current SOC state and driving condition of the vehicle; S220: when detecting that the current change rate of a single branch exceeds the baseline value by 150% for 5 seconds, activating an abnormal current event marker; S230: based on the physical position index of the abnormal current branch, locating the area coordinates of the corresponding battery module in the infrared thermal map; S240: performing dynamic contrast enhancement processing, and performing directional enhancement processing on the target area; S250: generating an abnormality degree score value of the thermal map after enhancement processing, and triggering a three-level alarm protocol when the score value exceeds a preset threshold. 4.The intelligent AI detection method based on infrared thermography and automotive diagnostics of claim 3, wherein, The step S300 specifically comprises the following steps: S310: based on the feature matching of the thermal map and a vehicle structure CAD model, dividing the thermal map into: a battery module area covering the battery pack shell contour and extending outward by 10% safety margin; a high-voltage cable area generating a variable-width strip-shaped detection area along the cable direction; an electronic control unit area being a rectangular detection frame with the geometric center of the motor controller shell as the reference; S320: performing differential gradient processing on each subarea; S330: dynamically adjusting the histogram equalization parameters according to the real-time ambient temperature: when the ambient temperature is greater than 35℃, limiting the gray scale stretching amplitude of the high-temperature area; when the ambient temperature is less than 0℃, enhancing the contrast sensitivity of the low-temperature area; S340: generating a subarea enhanced thermal map and calculating the contrast improvement coefficient of each area, and triggering an image reacquisition instruction when the coefficient does not reach the set standard.

5. The intelligent AI detection method based on infrared thermography and automotive diagnostics according to any one of claims 1-4, characterized in that, The step S400 specifically comprises the following steps: S410: dynamically adjust the temperature threshold of each partition based on real-time vehicle speed, ambient temperature and battery state of charge, wherein: The threshold of the battery module area increases with the increase of vehicle speed and decreases with the increase of ambient temperature; The threshold of the high-voltage wire harness area is dynamically corrected according to the charging current intensity; The threshold of the electronic control unit area is inversely related to the ambient temperature; S420: set different threshold update frequencies according to the vehicle driving or charging state; S430: when the temperature of a certain partition continuously exceeds the threshold, perform additional verification steps corresponding to the type of the partition.

6. The intelligent AI detection method based on infrared thermography and automotive diagnostics according to claim 5, characterized in that, The step S500 specifically comprises the following steps: S510: when the abnormal area of the thermal map and the abnormality of the electrical parameter are detected, obtain the component material parameters and adjacent sensor data of the area; S520: based on the surface temperature distribution of the abnormal area, establish a three-dimensional heat conduction model and simulate the internal heat field propagation path; S530: generate a thermal risk diffusion trend chart according to the simulation results, and mark the location of the internal potential overheating point; S540: compare the simulation results with the historical fault feature library, and output the fault type and risk level with the highest matching degree.

7. The intelligent AI detection method based on infrared thermography and automotive diagnostics according to claim 6, characterized in that, The step S600 specifically comprises the following steps: S610: extract the current abnormal features, including temperature distribution pattern, electrical parameter change curve and heat conduction trend; S620: perform similarity matching of the abnormal features with the historical fault case library, and calculate the matching scores of each case; S630: select the top three cases with the highest matching scores, assign weights according to the feature importance to generate a comprehensive risk level; S640: based on the fault location data of the matching cases, generate a visual positioning probability chart and mark the high-risk area; S650: store the current diagnosis results as a new case in the historical library, and optimize the subsequent matching priority. 8.The intelligent AI detection method based on infrared thermography and automotive diagnostics of claim 2, wherein, The step S140 specifically comprises the following steps: S141: add a unified timestamp to all data streams based on the vehicle clock source; S142: spatially register the infrared image coordinate system with the physical position of the vehicle components; S143: establish a mapping relationship matrix between the change of electrical parameters and the change of thermal map area. 9.The intelligent AI detection method based on infrared thermography and automotive diagnostics of claim 3, wherein, The step S240 specifically comprises the following steps: S241: use contrast adjustment to highlight the temperature difference details of the target area; S242: apply directional filtering processing along the extension direction of the battery module tab to strengthen the overheating features of the connecting piece area; S243: dynamically adjust the image enhancement intensity parameters according to the ambient temperature. 10.The intelligent AI detection method based on infrared thermography and vehicle diagnosis of claim 4, wherein, The step S320 specifically comprises the following steps: S321: the battery module area uses horizontal gradient enhancement to highlight the heat conduction features of the connecting piece between battery monomers; S322: the high-voltage wire harness area implements vertical filtering processing to strengthen the detection of cable axial temperature distribution abnormalities; S323: the electronic control unit area performs isotropic gradient operation to capture local overheating points of components.

Images