A multi-dimensional security risk early warning method and system

Abstract

This application discloses a multi-dimensional safety risk early warning method and system. The method includes: acquiring multi-dimensional safety data of a vehicle; inputting the data into a neural network model to predict the deterioration trend of key parameters; performing a dual-condition judgment based on the current absolute value and rate of change of the key parameters; and determining the early warning level and executing response actions based on the predicted trend and judgment results. The system includes a data acquisition unit, a central processing unit, and a risk execution unit to implement the method. This application can predict risks in advance and perform intelligent graded responses, thereby improving the sensitivity and accuracy of early warnings and the active safety of vehicles.

Description

Technical Field

[0001] This application belongs to the field of vehicle safety monitoring technology, specifically involving a multi-dimensional safety risk early warning method and system. Background Technology

[0002] Vehicles, especially RVs integrating complex living facilities, contain multiple safety elements, including driving systems, environmental systems, and energy systems. Traditional vehicle safety alarm systems typically employ an "over-limit alarm" mode based on fixed thresholds. For example, when the battery voltage drops below a preset 10V, the system triggers a buzzer alarm. This mode has significant technical drawbacks: First, delayed response. Alarms are usually triggered only when risk parameters have already reached a critical danger point, by which time irreversible damage to vehicle equipment may have already occurred, or even personal safety may be endangered. Second, limited functionality. Such systems cannot predict the development trend of risks and can only provide post-event warnings, lacking foresight and proactive defense capabilities. Third, high false alarm rate. For example, the instantaneous high current generated when starting a high-power inductive load (such as an air conditioning compressor), or the voltage drop caused by the battery's increased internal resistance due to normal aging, may be misjudged as a fault, thus frequently triggering false alarms, reducing system reliability and user experience.

[0003] Therefore, how to overcome the problems of delayed response, limited functionality and high false alarm rate of existing alarm systems, and provide an active early warning solution that can predict risk trends in advance and respond intelligently and differently according to risk levels is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0004] This application provides a multi-dimensional safety risk early warning method, system, electronic device, and computer-readable storage medium, which solves the problems of slow response, inability to predict risk trends, and high false alarm rate of existing vehicle alarm systems.

[0005] To achieve the above objectives, the first aspect of this application provides a multi-dimensional security risk early warning method, including: Acquire at least one piece of vehicle driving safety data, at least one piece of environmental safety data, and at least one piece of energy supply data to form multi-dimensional data; The multi-dimensional data is input into a pre-trained neural network model, which then predicts the deterioration trend of key parameters corresponding to the data within a preset time period in the future. The system performs a dual-condition judgment based on whether the current absolute value of the key parameter triggers a first threshold and whether the rate of change of the key parameter triggers a second threshold. When either condition is met, it is determined to trigger an early warning condition. Based on the deteriorating trend and / or the triggered warning conditions, determine the risk warning level and execute the corresponding preset response action according to the warning level.

[0006] The second aspect of this application provides a multi-dimensional security risk early warning system, including: The data acquisition unit is used to acquire at least one piece of driving safety data, at least one piece of environmental safety data, and at least one piece of energy supply data of the vehicle, forming multi-dimensional data; A central processing unit, electrically connected to the data acquisition unit, has a pre-trained neural network model built in it and is configured to: input the multi-dimensional data acquired by the data acquisition unit into the neural network model to predict the deterioration trend of key parameters corresponding to the data within a future preset time period; perform a dual condition determination based on whether the current absolute value of the key parameter triggers a first threshold and whether the rate of change of the key parameter triggers a second threshold, and determine that a warning condition is triggered when either condition is met; and determine the warning level of the risk based on the deterioration trend and / or the triggered warning condition. The risk execution unit is electrically connected to the central processing unit and is used to execute a corresponding preset response action based on the warning level determined by the central processing unit.

[0007] This application also provides an electronic device configured to perform the steps of the above-described method.

[0008] This application also provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the above-described method.

[0009] This application has the following beneficial effects: 1. By introducing a neural network model to analyze multi-dimensional data, it is possible to predict the future deterioration trend of key parameters, realizing the transformation from "post-event alarm" to "pre-event warning", and buying time for risk disposal.

[0010] 2. By adopting a dual-condition judgment mechanism of "absolute value-rate of change", it not only focuses on the static value of the parameter, but also on its dynamic changes, which can identify early signs of sudden risks such as fire and short circuit more early and effectively.

[0011] 3. Based on the predicted risk level, differentiated response actions are executed, ranging from simple information prompts to mandatory safety measures such as automatic power cut-off. This achieves intelligent risk management, enabling automatic intervention in emergencies to prevent the situation from escalating and significantly improving the overall safety of the vehicle.

[0012] 4. By combining waveform fingerprint matching technology and adaptive threshold adjustment technology based on aging factors, it is possible to effectively distinguish between normal load startup and real faults, and compensate for parameter drift caused by equipment aging, thereby significantly reducing the false alarm rate and improving the reliability and practicality of the early warning system. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application 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 this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a flowchart illustrating a multi-dimensional security risk early warning method provided in an embodiment of the present invention.

[0015] Figure 2 This is a functional module block diagram of a multi-dimensional security risk early warning system provided in an embodiment of the present invention.

[0016] Figure 3 This is a schematic diagram of the heterogeneous dual-core hardware architecture of the central processing unit provided in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] It should be noted that in the description of this application, the terms "connection," "installation," and "setting," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0019] Example 1 This embodiment provides a multi-dimensional safety risk early warning method, which uses artificial intelligence technology to achieve predictive analysis and intelligent response to potential vehicle risks. See also Figure 1 This is a schematic diagram of the method flow provided in an embodiment of the present invention, which can be applied to... Figure 2 The system shown includes the following steps: Acquiring multi-dimensional data. The system uses multiple sensors deployed inside the vehicle to acquire multi-dimensional data reflecting the vehicle's operating status in real time. This data is divided into at least three categories: 1) Driving safety data, which reflects the safety status of the vehicle during driving, may include, for example, at least one of vehicle speed data and tire pressure data.

[0020] 2) Environmental safety data, which reflects the safety status of the environment in which the occupants are located, may include, for example, at least one of carbon monoxide concentration data and combustible gas concentration data in the vehicle.

[0021] 3) Energy supply data, which reflects the stability and safety of the vehicle's energy system. For example, it may include battery pack voltage data, current data, battery state of charge, and inverter operating power.

[0022] In this embodiment, the data acquisition unit converts the analog signals collected by the sensor into digital signals using an analog-to-digital converter for subsequent processing. For example, a high-frequency Hall current sensor is used to collect current data at a sampling frequency of 20 kHz to ensure that transient details can be captured.

[0023] AI trend prediction is performed. The acquired multi-dimensional data is input into a pre-trained neural network model. The core task of this model is to analyze the correlation between current data and historical data, and predict the "deteriorating trend" of key parameters within a preset time period (e.g., the next 5 or 30 minutes).

[0024] In a preferred embodiment, the neural network model is a Long Short-Term Memory (LSTM) network model. LSTM network models are particularly suitable for processing and predicting important events in time-series data, effectively learning long-term dependencies between data points. For example, by analyzing current discharge current and ambient temperature data, an LSTM network model can predict the future rate of decline in battery state of charge. Alternatively, by analyzing subtle trends in carbon monoxide concentration inside the vehicle, the model can provide early warnings of potential exhaust gas leaks. As an alternative, other types of neural networks, such as backpropagation (BP) neural networks, can also be used.

[0025] A dual-condition judgment mechanism is implemented. To achieve highly sensitive risk identification, this method employs an "absolute value-rate of change" dual-condition judgment mechanism. This mechanism sets two parallel trigger conditions for key parameters: 1. Absolute value determination: This determines whether the current absolute value of a key parameter has triggered a preset first threshold. For example, the absolute value of the battery voltage is below 10V. This is a traditional, fallback static safety boundary.

[0026] 2. Rate of Change Detection: This function determines whether the rate of change of a key parameter (e.g., the first derivative) over a unit of time has triggered a preset second threshold. For example, even if the battery voltage is currently 12V and has not yet triggered the first threshold, a voltage drop rate exceeding 0.5V / s within 1 second may indicate an impending short circuit or load anomaly.

[0027] When any of the above conditions is met, the system determines that the warning condition has been triggered. This dual-determination mechanism can detect the early signs of sudden failures earlier than a single absolute value determination.

[0028] The system determines the warning level and executes response actions. Once a warning condition is triggered, the system combines the AI ​​prediction results from the previous step with / or the specific triggering conditions to comprehensively assess the severity of the risk and map it to a preset warning level. In this embodiment, the warning level is divided into four levels: *Level III (Severe) warning level: This indicates the highest level of risk, which could result in equipment damage or personal injury if not addressed immediately. It is typically indicated on the vehicle's central touchscreen by an icon consisting of a red triangle with an exclamation mark inside.

[0029] * Warning level is Level II (moderate): indicated by an orange triangle icon.

[0030] *The warning level is Level I (general level): indicated by a "yellow triangle" icon.

[0031] * Warning level 0 (advice level): indicated by a "gray triangle" icon.

[0032] Based on the determined warning level, the system will automatically execute a pre-set response action that matches it. For example: * For Level I "Battery Low Voltage Protection Warning", the system may only pop up a yellow warning window on the central control screen and provide a troubleshooting guide to prompt early manual intervention.

[0033] * For Level III “Main Power Disconnection Risk” or “Severe Short Circuit Warning”, the system will immediately trigger a red severe warning and automatically cut off the associated main power supply through the risk execution unit (such as a relay array) to prevent equipment burnout or fire.

[0034] To further improve the accuracy of early warning and reduce the false alarm rate, this method may optionally include the following two optimization steps: Transient current suppression based on waveform fingerprinting. During vehicle operation, the startup of high-power inductive loads such as air conditioners, microwave ovens, and water pumps generates huge transient inrush currents. The rate of change of these currents may trigger the second threshold in the previous step, leading to false alarms. To address this issue, a transient current suppression step is introduced before performing the dual-condition judgment.

[0035] The system pre-establishes a fingerprint database, which stores standard current waveforms generated by common electrical devices (such as air conditioners and microwave ovens) during normal startup. These standard waveforms are called "characteristic fingerprints".

[0036] When the system detects a transient current event, it will collect the waveform data of the transient current in real time (for example, collect the waveform for 200ms).

[0037] The system calculates the "morphological distance" between the real-time waveform and each feature fingerprint stored in the fingerprint database. If the calculated minimum morphological distance is less than a preset distance threshold... If the transient current event is detected, the system determines that it is a legitimate, known load initiation event, rather than a fault. In this case, the system will suppress any warnings that might be triggered by this transient current and only record the event log, thus effectively avoiding false alarms.

[0038] Specifically, the preferred algorithm for calculating morphological distance is the Dynamic Time Warping (DTW) algorithm. DTW can effectively measure the similarity between two time series of potentially different lengths. The system uses DTW to calculate real-time waveform sequences. Compared with the fingerprint sequences of various features in the fingerprint database Distance between This calculation logic can be implemented using the following formula: in, This represents the optimal normalized path between two time series. This is the Euclidean distance between corresponding data points along the path. The minimum time series distance calculated at the end is considered the morphological distance.

[0039] If the calculated minimum morphological distance Greater than or equal to the distance threshold and the rate of change of current Exceeded the danger threshold If this occurs, the system will determine it as an "abnormal mutation" and immediately trigger a Level III alarm.

[0040] Using this method, the system can reduce the false alarm rate caused by the start-up of high-power inductive loads to below 0.1% while ensuring that the response time to real short-circuit faults is less than 20ms.

[0041] Adaptive Threshold Adjustment Based on Aging Factors. Vehicle energy storage components (such as lithium iron phosphate battery packs) age over time, a significant characteristic being an increase in DC internal resistance. Increased internal resistance leads to a larger internal voltage drop under the same load, potentially causing the voltage during normal operation to fall below a fixed first threshold, thus triggering false alarms. To address this issue, this method introduces an adaptive threshold mechanism.

[0042] The system sets the first threshold as an adaptive threshold.

[0043] The system periodically (e.g., every 24 hours) applies a microsecond-level pulse load when the vehicle is under no-load or low-load conditions, and accurately calculates the health status parameters of the energy storage component by measuring changes in voltage and current. In this embodiment, the energy storage component is a battery pack, and the health status parameter is the DC internal resistance of the battery pack. This DC internal resistance value is quantified as an "aging factor" characterizing the battery's health status.

[0044] The system dynamically adjusts the specific value of the voltage-related adaptive threshold based on the calculated aging factor to compensate for parameter baseline drift caused by battery performance degradation. Specifically, the system calculates the voltage compensation amount based on the difference between the currently measured DC internal resistance and the battery's factory reference internal resistance value. Subsequently, the voltage-related first threshold is subtracted from its reference value to obtain the adjusted new threshold suitable for the current battery health state.

[0045] In a preferred embodiment, the dynamic adjustment logic of the alarm threshold is implemented through the following formula: in, It is a dynamically adjusted real-time alarm threshold. This is the reference alarm voltage for the battery under ideal conditions. This is the current discharge current. This is the latest calculated DC internal resistance of the battery. It is a safety margin factor (for example, a value of 1.2) used to provide additional safety buffers.

[0046] This model is used when the battery ages... When the voltage is increased from 5mΩ to 10mΩ, the system will automatically lower the undervoltage warning trigger threshold. This avoids false alarms caused by increased internal resistance voltage, ensuring the effectiveness and accuracy of the warning system throughout the battery's entire lifespan.

[0047] Example 2 This embodiment provides a multi-dimensional security risk early warning system, which is the hardware carrier of the method described in Embodiment 1. See also... Figure 2 This is a functional block diagram of the system, which includes a data acquisition unit 201, a central processing unit 202, and a risk execution unit 203.

[0048] The data acquisition unit 201 is responsible for performing the data acquisition steps in the above method. It consists of an array of sensors distributed throughout the vehicle, such as Hall current sensors, voltage sampling circuits, temperature sensors, gas concentration sensors, and tire pressure sensors. This unit also includes necessary signal conditioning circuitry and an analog-to-digital converter to convert the acquired physical signals into digital signals that can be processed by the central processing unit 202.

[0049] The central processing unit 202, which is the "brain" of the entire system, is electrically connected to the data acquisition unit 201. It is responsible for receiving and processing data, performing core complex calculations such as AI prediction, dual-condition judgment, and risk level assessment. The central processing unit 202 internally contains or is loaded with a pre-trained neural network model (such as a long short-term memory network model) and is configured to execute the steps described in Embodiment 1 and the logic for determining the warning level.

[0050] The risk execution unit 203, acting as the "arm" of the system, is electrically connected to the central processing unit 202. It is responsible for receiving instructions from the central processing unit 202 and executing specific response actions. This unit may include various actuators, such as a human-machine interface driver circuit for displaying alarm information on the central control screen, a buzzer and warning lights for emitting audible and visual alarms, and a relay or solid-state switch array for cutting off circuits.

[0051] In a specific, preferred embodiment, see [link to specific embodiment]. Figure 3 The central processing unit 202 employs a heterogeneous dual-core hardware architecture to achieve a balance between high performance and low power consumption. This architecture includes: The field-programmable gate array (FPGA) preprocessing unit 301, with its parallel processing and high real-time performance, serves as the system's "sentinel." It is directly connected to high-frequency sensors (such as current and voltage sensors) in the data acquisition unit 201 and configured to perform real-time preprocessing of high-frequency data with sampling frequencies up to 20kHz. Specifically, it internally incorporates differential calculation logic for real-time calculation of the rate of change of key parameters (such as the rate of change of current). It then compares the result with a second threshold set by the hardware (e.g., 50 A / ms). This hardware-level comparison is extremely fast and consumes very little power.

[0052] The ARM Master Control Unit 302 is typically a high-performance ARM Cortex-A series processor running an embedded Linux operating system, acting as the system's "commander." It is responsible for handling complex, non-real-time computing tasks.

[0053] The field-programmable gate array (FPGA) preprocessing unit 301 and the ARM main control unit 302 are connected via a high-speed SPI (Serial Peripheral Interface) bus. Additionally, the system acquires vehicle chassis data via a CAN (Controller Area Network) bus. Normally, the ARM main control unit 302 can be in a low-power or sleep state. Only when the FPGA preprocessing unit 301 detects a sudden change in the parameter rate of change and triggers the second threshold will it immediately wake up the ARM main control unit 302 via an interrupt signal and transmit the cached detailed waveform data (e.g., 200ms) to the ARM main control unit 302.

[0054] After being woken up, the ARM main control unit 302 is configured to perform a higher-level comprehensive risk assessment, including: running a long short-term memory network neural network model stored in non-volatile memory to predict deterioration trends; executing the waveform morphological matching algorithm described in Embodiment 1 to suppress transient false alarms; and performing adaptive threshold calculation based on aging factors. Finally, the ARM main control unit 302 makes a final risk level determination and sends instructions to the central control screen and power management module in the risk execution unit 203 via the CAN bus to execute corresponding response actions.

[0055] This heterogeneous architecture, which combines a field-programmable gate array (FPGA) with an ARM processor, achieves an optimal balance between system performance, power consumption, and response speed. The FPGA is used to process high-frequency, simple real-time tasks, while the ARM processor is used to process low-frequency, complex intelligent algorithms.

[0056] In addition, this application also provides an electronic device including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in Embodiment 1.

[0057] This application also provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method described in Embodiment 1.

[0058] In summary, this application constructs a core framework of multi-dimensional data collection, AI trend prediction, dual-condition judgment, and hierarchical response, and innovatively incorporates transient suppression technology based on waveform fingerprinting and adaptive threshold technology based on aging factors. This not only enables early warning and intelligent control of vehicle safety risks, but also significantly improves the accuracy of warnings and adaptability throughout the entire life cycle, thereby solving the key problems existing in the background technology.

[0059] It should be noted that the above embodiments are merely preferred embodiments of this application, intended to illustrate the technical solutions of this application, and not to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A multi-dimensional security risk early warning method, characterized in that, include: Acquire at least one piece of vehicle driving safety data, at least one piece of environmental safety data, and at least one piece of energy supply data to form multi-dimensional data; The multi-dimensional data is input into a pre-trained neural network model, which then predicts the deterioration trend of key parameters corresponding to the data within a preset time period in the future. The system performs a dual-condition judgment based on whether the current absolute value of the key parameter triggers a first threshold and whether the rate of change of the key parameter triggers a second threshold. When either condition is met, it is determined to trigger an early warning condition. Based on the deteriorating trend and / or the triggered warning conditions, determine the risk warning level and execute the corresponding preset response action according to the warning level.

2. The method according to claim 1, characterized in that, The driving safety data includes at least one of vehicle speed data and tire pressure data; the environmental safety data includes at least one of in-vehicle carbon monoxide concentration data and combustible gas concentration data; the energy supply data includes at least one of battery pack voltage data and current data.

3. The method according to claim 1, characterized in that, The neural network model is a long short-term memory network model.

4. The method according to claim 1, characterized in that, Prior to the dual-condition determination, the method further includes a transient current suppression step, which includes: A fingerprint database is pre-stored, using the current waveform of electrical equipment during normal startup as a feature fingerprint. When a transient current is detected, the real-time waveform of the transient current is acquired; Calculate the morphological distance between the real-time waveform and each of the characteristic fingerprints in the fingerprint database; If the calculated morphological distance is less than a preset distance threshold, the current transient current is determined to be a legitimate load initiation event, and any warnings that may be triggered by the transient current are suppressed.

5. The method according to claim 4, characterized in that, The step of calculating the morphological distance between the real-time waveform and each of the feature fingerprints in the fingerprint database specifically involves: using a dynamic time warping algorithm to calculate the time series distance between the real-time waveform and each of the feature fingerprints, and taking the smallest calculated time series distance as the morphological distance.

6. The method according to claim 1, characterized in that, The first threshold is an adaptive threshold; the method further includes: The health status parameters of energy storage components in the vehicle energy system are periodically measured, and the health status parameters are quantified as aging factors. The value of the adaptive threshold is dynamically adjusted based on the aging factor to compensate for the parameter baseline drift caused by the performance degradation of the energy storage component.

7. The method according to claim 6, characterized in that, The energy storage component is a battery pack, and the health status parameter is the DC internal resistance of the battery pack; the step of dynamically adjusting the adaptive threshold according to the aging factor specifically includes: The voltage compensation amount is calculated based on the difference between the measured value of the DC internal resistance and the reference internal resistance value. The adjusted first threshold is obtained by subtracting the voltage compensation amount from its reference value from the first threshold related to voltage.

8. A multi-dimensional security risk early warning system, characterized in that, include: The data acquisition unit is used to acquire at least one piece of driving safety data, at least one piece of environmental safety data, and at least one piece of energy supply data of the vehicle, forming multi-dimensional data; A central processing unit is electrically connected to the data acquisition unit. The central processing unit has a pre-trained neural network model built in and is configured to: input the multi-dimensional data acquired by the data acquisition unit into the neural network model in order to predict the deterioration trend of the key parameters corresponding to the data within a future preset time period. The system performs a dual-condition judgment based on whether the current absolute value of the key parameter triggers a first threshold and whether the rate of change of the key parameter triggers a second threshold. When either condition is met, it is determined to trigger an early warning condition. And based on the aforementioned deterioration trend and / or the triggered warning conditions, determine the risk warning level; The risk execution unit is electrically connected to the central processing unit and is used to execute a corresponding preset response action based on the warning level determined by the central processing unit.

9. The system according to claim 8, characterized in that, The central processing unit 202 adopts a heterogeneous dual-core architecture, including: Field Programmable Gate Array (FPGA) preprocessing unit 301; and ARM (Advanced RISC Machine) main control unit 302.

10. The system according to claim 9, characterized in that, The field-programmable gate array preprocessing unit is configured to perform real-time preprocessing on the high-frequency sampling data collected by the data acquisition unit in order to detect whether the rate of change of the key parameter changes abruptly and determine whether the second threshold is triggered. The ARM main control unit is configured to run the neural network model to predict the deterioration trend and perform a comprehensive risk assessment after receiving a trigger signal from the field programmable gate array preprocessing unit.

Images