Automobile exhaust flow real-time monitoring system based on intelligent sensor

Abstract

The invention discloses an automobile exhaust flow real-time monitoring system based on an intelligent sensor, and the system comprises a multi-mode sensor array which is connected with an edge calculation unit and is used for collecting automobile exhaust parameters; the edge computing unit is connected with the cloud monitoring platform and is used for processing the parameters acquired by the multi-modal sensor array; and the cloud monitoring platform establishes a multi-physics field coupling model through a cloud platform to realize emission trend prediction and abnormity alarm. The automobile exhaust flow monitoring system has the advantages of being capable of accurately monitoring the automobile exhaust flow in real time, high in precision, high in reliability, intelligent, wide in applicability and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of automotive exhaust emission monitoring. Specifically, it particularly relates to a real-time monitoring system for automotive exhaust gas flow based on intelligent sensors. Background Art

[0002] With the increasing severity of global environmental problems, the impact of automotive exhaust emissions on air quality has received more and more attention. Automotive exhaust contains various harmful substances, such as carbon monoxide, carbon dioxide, nitrogen oxides, hydrocarbons, and particulate matter, etc. These substances pose a serious threat to the environment and human health. In order to reduce the negative impact of automotive exhaust emissions on the environment;

[0003] The existing methods for automotive exhaust emission monitoring mainly rely on regular inspections and laboratory analyses. Although these methods can provide relatively accurate emission data, they have problems such as poor real-time performance, high cost, and complex operation. With the development of the Internet of Things, sensor technology, and edge computing, it has become possible to monitor automotive exhaust emissions in real time. By deploying an intelligent sensor array, automotive exhaust parameters can be collected in real time, and combined with edge computing and cloud platforms for data analysis and processing, so as to achieve emission trend prediction and abnormal alarm;

[0004] However, on the one hand, in the existing automotive exhaust monitoring systems, the automotive exhaust environment is complex and changeable, and sensors are prone to measurement errors under harsh conditions such as high temperature, high humidity, and vibration, affecting the accuracy of data. On the other hand, the data collected by multi-modal sensors need to be synchronized and integrated, and traditional time synchronization and data integration methods are difficult to meet the high-precision requirements in a dynamic exhaust environment. On the other hand, sensor failures or environmental interferences may cause data anomalies. How to monitor data quality in real time, detect sensor status, and process abnormal data is also a problem that needs to be solved by the existing technology.

[0005] Therefore, those skilled in the art are committed to providing a real-time monitoring system for automotive exhaust gas flow based on intelligent sensors, which can overcome the deficiencies of the existing technology, achieve real-time and accurate monitoring of automotive exhaust parameters, and perform emission trend prediction and abnormal alarm through a multi-physical field coupling model. Summary of the Invention

[0006] In view of the above-mentioned defects of the existing technology, the technical problem to be solved by the present invention is to provide a real-time monitoring system for automotive exhaust gas flow based on intelligent sensors that can effectively solve the above-mentioned technical problems.

[0007] To achieve the above object, the present invention provides a real-time monitoring system for automotive exhaust gas flow based on intelligent sensors, including

[0008] A multi-modal sensor array, connected to an edge computing unit, for collecting parameters of vehicle exhaust;

[0009] An edge computing unit, connected to a cloud monitoring platform, for processing the parameters collected by the multi-modal sensor array;

[0010] A cloud monitoring platform, establishing a multi-physical field coupling model through the cloud platform to achieve emission trend prediction and abnormal alarm.

[0011] Furthermore, the specific steps for the multi-modal sensor array to collect parameters of vehicle exhaust include:

[0012] S1: Determine the vehicle exhaust parameters to be collected;

[0013] S2: Select multi-modal sensors according to the determined vehicle exhaust parameters to be collected;

[0014] S3: Set the sensor array selected in S2 at the vehicle exhaust pipe;

[0015] S4: Calibrate the multi-modal sensors;

[0016] S5: Start the multi-modal sensor array to collect exhaust parameters;

[0017] S6: Synchronize and integrate the data collected by the multi-modal sensors;

[0018] S7: Transmit the collected exhaust parameter data to the edge computing unit.

[0019] Furthermore, during the process of collecting exhaust parameters, it also includes monitoring data quality, detecting whether the sensors are working properly, and marking or eliminating abnormal data.

[0020] Furthermore, the specific calibration of the multi-modal sensors in S4 specifically includes:

[0021] S4B1: Collect data of the multi-modal sensors and reference devices in a dynamic exhaust environment, and compare and analyze the errors;

[0022] S4B2: Use redundant sensor data fusion to generate reference values and calibrate the output of the multi-modal sensors;

[0023] S4B3: Use historical data to train the calibration model and adjust the sensor output;

[0024] S4B4: Use the physical relationship between the multi-modal sensors for cross-calibration;

[0025] S4B5) During the vehicle operation, analyze the sensor data and dynamically adjust the calibration parameters.

[0026] Further, before the step S4B1, the following steps are also included:

[0027] S4A1: Define the parameters to be calibrated and their accuracy requirements;

[0028] S4A2: Prepare a reference sensor or instrument as the calibration benchmark;

[0029] S4A3: Build a dynamic exhaust environment simulation device in the laboratory or in an actual vehicle, which can simulate the exhaust conditions when the vehicle is running;

[0030] S4A4: Install the multi-modal sensor array in the exhaust pipe;

[0031] S4A5: Configure a data acquisition system for synchronously acquiring the data of the multi-modal sensor and the reference device.

[0032] Further, after the step S4B5, the following steps are also included:

[0033] S4C1: Dynamically adjust the calibration parameters of the sensor according to the change of environmental parameters;

[0034] S4C2: Record the calibration data in the blockchain;

[0035] S4C3: Verify the calibration effect in the actual operating environment;

[0036] S4C4: Regularly check the calibration status of the sensor, and re-calibrate or update the calibration model if necessary.

[0037] Further, the synchronization and integration of the data collected by the multi-modal sensor in the step S6) include:

[0038] Distributed time synchronization based on edge computing

[0039] S6A1: Deploy a distributed time synchronization algorithm in the edge computing unit;

[0040] S6A2: Assign a unique timestamp to each sensor node and uniformly calibrate the time through the edge computing unit;

[0041] S6A3: During the data acquisition process, the edge computing unit receives the data of each sensor in real time and aligns them according to the timestamps;

[0042] S6A4: Use interpolation or extrapolation methods to fill in the missing data caused by communication delays or different sampling rates;

[0043] Data synchronization based on event triggering

[0044] S6B1: Define several trigger events;

[0045] S6B2: When a certain sensor detects a trigger event, send a synchronization signal to the edge computing unit;

[0046] S6B3: After receiving the synchronization signal, the edge computing unit collects the data of other sensors and performs time alignment;

[0047] S6B4: Integrate the synchronized data to generate a unified data packet;

[0048] Data-driven intelligent synchronization

[0049] S6C1: Use machine learning algorithms to analyze the time series characteristics of multi-modal sensor data;

[0050] S6C2: Identify the time offset according to the data characteristics and perform dynamic alignment;

[0051] S6C3: Perform data correction during the data integration process;

[0052] S6C4: Store the synchronized and integrated data into a unified data structure;

[0053] Synchronization and integration based on multi-physical field coupling

[0054] S6D1: Establish a multi-physical field coupling model;

[0055] S6D2: During the data synchronization process, use the coupling model to predict the data change trends of each sensor;

[0056] S6D3: According to the prediction results, dynamically adjust the time alignment method of the sensor data;

[0057] S6D4: During the data integration process, map the multi-modal data into a unified physical field model to generate comprehensive data;

[0058] Data synchronization and integration based on blockchain

[0059] S6E1: Upload the data collected by each sensor to the blockchain network in real time;

[0060] S6E2: Use the consensus mechanism of the blockchain to ensure the time consistency and immutability of the data;

[0061] S6E3: In the blockchain network, perform smart contract-driven synchronization and integration on the multi-modal data;

[0062] S6E4: Store the integrated data into the blockchain;

[0063] Real-time synchronization based on adaptive filtering

[0064] S6F1: Deploy an adaptive filtering algorithm in the edge computing unit;

[0065] S6F2: Receive the data of each sensor in real time, and perform time alignment and data correction according to the filtering algorithm;

[0066] S6F3: Dynamically adjust the filtering parameters during the data integration process to adapt to the changes in sensor data;

[0067] S6F4: Output the synchronized and integrated data for real-time monitoring and analysis;

[0068] Heterogeneous data integration based on graph model

[0069] S6G1: Model the multi-modal sensor data as a graph structure, where nodes represent sensors and edges represent the relationships between data;

[0070] S6G2: Use graph neural networks to extract features and fuse heterogeneous data;

[0071] S6G3: Dynamically adjust the data synchronization and integration strategy according to the analysis results of the graph model;

[0072] S6G4: Output the integrated data for analysis and decision-making.

[0073] Furthermore, the marking of abnormal data includes abnormal data detection and abnormal data marking;

[0074] Abnormal data detection

[0075] a1) Set a numerical range for each sensor, and check whether the sensor data exceeds the range according to the set numerical range;

[0076] If it is found that the sensor data exceeds the range, trigger the alarm mechanism; at the same time, record and store the abnormal data;

[0077] If the sensor data does not exceed the range, continue normal monitoring, and continuously obtain and analyze the sensor data;

[0078] a2) Use a sliding window or differential method to detect sudden changes in data, and set a mutation threshold at the same time;

[0079] a3) Check whether the physical relationships between multi-modal sensor data are reasonable;

[0080] If the physical relationships are reasonable, integrate the multi-modal data into the final result through data fusion technology; extract features from the multi-modal data for subsequent machine learning or deep learning model training, for example, extract motion features from accelerometer and gyroscope data; use the multi-modal data to train a prediction model; monitor the physical relationships of the multi-modal data;

[0081] If the physical relationship is unreasonable, there may be sensor failures or calibration errors. Individually test or recalibrate the suspicious sensors; check for environmental interference causing data anomalies. If it is environmental interference, shield the interference source or adjust the sensor position; clean or interpolate the abnormal data;

[0082] a4) Check if the sensor signal is stable; use the sensor self-diagnostic function to detect hardware failures. If the sensor signal suddenly interrupts, mark it as a failure;

[0083] Abnormal data marking

[0084] b1) Add a field to each data record to mark whether the data is abnormal;

[0085] B2) Mark the type and degree of the anomaly, and record the time when the anomaly occurred.

[0086] Furthermore, after marking the abnormal data, it also includes eliminating the abnormal data, specifically including temporary elimination, data repair, and permanent elimination.

[0087] Furthermore, in S4B1, collect data from multi-modal sensors and reference devices in a dynamic exhaust environment. The comparative analysis of errors includes: data collection, data preprocessing, and error analysis;

[0088] The data collection includes building a unified synchronous triggering system for multi-modal sensors and reference devices; and while collecting data, record the parameters in the exhaust environment through the environmental monitoring module, so that the complete recording of environmental information helps to more comprehensively analyze the source of errors;

[0089] The data preprocessing includes preprocessing the collected data according to the recorded environmental parameters. And for the type data provided by the multi-modal sensors, extract the characteristic information of each modal data, and then align the data of different modalities in time series according to these characteristic information, so that when conducting comparative analysis, the multi-modal data and the reference device data are accurately matched in the time dimension;

[0090] The error analysis includes using machine learning algorithms to train the collected data to establish an error pattern recognition model; in a dynamic exhaust environment, if the error propagates and accumulates in the data sequence over time, then use the method of dynamic system modeling to analyze the propagation law of the error between different times and different measurement points, and predict the development trend of the error by establishing an error propagation model.

[0091] The beneficial effects of the present invention are:

[0092] 1. Real-time collect automotive exhaust parameters through a multi-modal sensor array to ensure the timeliness and accuracy of data; the sensor array can work stably in harsh environments such as high temperature, high humidity, and vibration to ensure the reliability of data;

[0093] 2. The system can collect data and analyze errors in a dynamic exhaust environment, adapting to the complex exhaust conditions during vehicle operation; through technologies such as dynamic time warping and multi-physical field coupling models, ensure the synchronization and integration accuracy of sensor data in a dynamic environment;

[0094] 3. Use redundant sensor data fusion, machine learning models, and historical data for intelligent calibration, dynamically adjust sensor output, reduce measurement errors, and ensure the consistency and accuracy of data through cross-calibration and the physical relationships between multi-modal sensors;

[0095] 4. Adopt various synchronization technologies such as distributed time synchronization, event-triggered synchronization, and data-driven synchronization based on edge computing to ensure the time consistency of multi-modal sensor data, and through technologies such as multi-physical field coupling models and graph neural networks, achieve intelligent integration of heterogeneous data, improving the accuracy and efficiency of data analysis.

[0096] 5. Real-time monitor data quality, detect whether the sensors are working properly, avoid data anomalies caused by sensor failures or environmental interference, mark, eliminate, or repair abnormal data to ensure the accuracy and reliability of data.

[0097] 6. Establish a multi-physical field coupling model through a cloud monitoring platform to achieve emission trend prediction and anomaly alarm, and use blockchain technology to record calibration data and error analysis results to ensure the immutability and traceability of data.

[0098] 7. This system is not only applicable to automotive exhaust monitoring but can also be extended to other fields such as industrial monitoring, the Internet of Things, and intelligent transportation, with broad application prospects; through modular design, the system can be expanded and customized according to actual needs to meet the monitoring requirements in different scenarios;

[0099] In summary, the present invention has the beneficial effects of real-time and accurate monitoring of automotive exhaust flow, with high precision, high reliability, intelligence, and broad applicability. Brief Description of the Drawings

[0100] Figure 1 It is a schematic flow diagram of a real-time automotive exhaust flow monitoring system based on intelligent sensors. Detailed Embodiment

[0101] The following further describes the present invention in conjunction with the drawings and embodiments:

[0102] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention. In addition, the terms "first", "second", "third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0103] In the description of the present invention, it should be noted that unless otherwise clearly specified and limited, the terms "installed", "set", "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0104] As Figure 1 shown

[0105] An in - vehicle exhaust gas flow rate real - time monitoring system based on intelligent sensors, comprising

[0106] S100: A multi - modal sensor array, connected to an edge computing unit, for collecting parameters of vehicle exhaust gas;

[0107] S200: An edge computing unit, connected to a cloud monitoring platform, for processing the parameters collected by the multi - modal sensor array;

[0108] S300: A cloud monitoring platform, which establishes a multi - physical - field coupling model through the cloud platform to achieve emission trend prediction and abnormal alarm.

[0109] In the present invention, the specific steps for the multi - modal sensor array to collect parameters of vehicle exhaust gas include:

[0110] S1: Determine the vehicle exhaust gas parameters to be collected; (selected according to actual needs) In the present invention, the vehicle exhaust gas parameters include but are not limited to the following parameters:

[0111] Gas concentration: such as CO (carbon monoxide), CO 2 (carbon dioxide), NOx (nitrogen oxides), HC (hydrocarbons), etc.; Physical parameters: such as exhaust gas temperature, exhaust gas pressure, exhaust gas flow rate, etc. Particle concentration: such as the content of PM2.5, PM10 and other particles.

[0112] S2: Select a multi-modal sensor according to the determined automotive exhaust parameters to be collected; including gas sensors: used to detect the concentration of various gases in the exhaust (such as electrochemical sensors, infrared sensors, etc.); temperature sensors: used to measure the exhaust temperature (such as thermocouples, thermistors, etc.); pressure sensors: used to measure the exhaust pressure (such as piezoresistive sensors); flow sensors: used to measure the exhaust flow rate (such as thermal flow meters, ultrasonic flow meters, etc.); particulate sensors: used to detect the concentration of particulate matter in the exhaust (such as laser scattering sensors).

[0113] S3: Set the sensor array selected in S2 at the automotive exhaust pipe; further, the set position ensures that the sensors can directly contact the exhaust gas flow to avoid interference. At the same time, avoid mutual interference between sensors; in addition, fix and protect the sensors to ensure their stable operation in harsh environments such as high temperature, high humidity, and vibration;

[0114] Specific high-temperature protection includes the following aspects. On the one hand, a high-temperature-resistant sensor housing, ceramic or special alloy, can be selected, which can withstand the high temperature of the exhaust (usually up to above 500°C), and at the same time, heat dissipation holes are opened on the sensor; on the other hand, a heat insulation layer (such as ceramic fiber or asbestos) can be installed between the sensor and the exhaust pipe to reduce the direct conduction of heat to the inside of the sensor;

[0115] High-humidity protection includes using shock-absorbing brackets or rubber pads to fix the sensors to reduce the impact of vibrations generated during vehicle driving on the sensors; using shock-absorbing materials (such as silicone pads) inside the sensors to fix the sensitive elements to reduce the impact of vibrations on the measurement accuracy.

[0116] S4: Calibrate the multi-modal sensor; specifically, calibrate the gas sensor using a standard gas or a gas sample with a known concentration; calibrate the temperature sensor and the pressure sensor using a standard temperature source and a pressure source; ensure the measurement accuracy and consistency of each sensor.

[0117] S5: Start the multi-modal sensor array to collect exhaust parameters; in the present invention, the specifically collected parameters include: gas concentration: the concentration of each gas in the exhaust is detected in real time by the gas sensor; temperature: the exhaust temperature is measured in real time by the temperature sensor; pressure: the exhaust pressure is measured in real time by the pressure sensor; flow rate: the exhaust flow rate is measured in real time by the flow sensor; particulate concentration: the content of particulate matter in the exhaust is detected in real time by the particulate sensor.

[0118] S6: Synchronize and integrate the data collected by the multi-modal sensor; specifically, use a unified time stamp to mark the data of each sensor to ensure the time consistency of the data; integrate the data of different sensors into a unified data format for subsequent processing and analysis.

[0119] Synchronizing and integrating the data collected by multi-modal sensors is a crucial step to ensure data consistency and improve analysis accuracy. Furthermore, it also includes:

[0120] Distributed time synchronization based on edge computing

[0121] S6A1: Deploy a distributed time synchronization algorithm in the edge computing unit;

[0122] S6A2: Assign a unique timestamp to each sensor node and uniformly calibrate the time through the edge computing unit;

[0123] S6A3: During the data acquisition process, the edge computing unit receives the data of each sensor in real time and aligns them according to the timestamps;

[0124] S6A4: Use interpolation or extrapolation methods to fill in the missing data caused by communication delays or different sampling rates;

[0125] Through the above steps, the present invention achieves high-precision time synchronization and reduces hardware costs; it is applicable to data acquisition in multi-sensor and high-dynamic environments.

[0126] Event-triggered data synchronization

[0127] S6B1: Define several trigger events (such as the exhaust gas temperature reaching a certain threshold, sudden change in gas concentration, etc.);

[0128] S6B2: When a certain sensor detects a trigger event, send a synchronization signal to the edge computing unit;

[0129] S6B3: After receiving the synchronization signal, the edge computing unit collects the data of other sensors and performs time alignment;

[0130] S6B4: Integrate the synchronized data to generate a unified data packet.

[0131] Through the above steps, the present invention reduces unnecessary synchronization operations, reduces the computational load, is applicable to event-driven monitoring scenarios, and improves data correlation;

[0132] Data-driven intelligent synchronization

[0133] S6C1: Use machine learning algorithms to analyze the time series characteristics of multi-modal sensor data. The formula for the dynamic time warping algorithm is,

[0134]

[0135] where DTW measures the similarity or distance between two time series;

[0136] Among them, X and Y respectively represent two time series to be compared;

[0137] π represents the set of all possible sequence alignment paths;

[0138] (i, j) ∈ π means traversing each element pair in these paths;

[0139] d(xi, yj) represents the distance (such as the Euclidean distance, etc.) between the ith element in the time series X and the jth element in the time series Y;

[0140] represents the summation of the distances of all element pairs on the selected path;

[0141] represents selecting the path with the minimum total distance among all the above possible paths and calculating its total distance as the DTW distance between the two time series X and Y;

[0142] It is used for time alignment of multi-modal sensor data to ensure that in a dynamic exhaust environment, the time series of sensor data can accurately match the reference device data; by minimizing the distance between time series, the time offset of sensor data is dynamically adjusted to reduce errors caused by time asynchronization.

[0143] S6C2: Automatically identify the time offset according to data characteristics and perform dynamic alignment;

[0144] S6C3: During data integration, consider the physical relationships between sensors (such as the influence of temperature on gas concentration) and perform data correction;

[0145] S6C4: Store the synchronized and integrated data into a unified data structure for subsequent analysis.

[0146] Through the above steps, the present invention automatically adapts to the time offset of sensor data, reduces manual intervention, and improves the accuracy and flexibility of data integration.

[0147] Synchronization and integration based on multi-physical field coupling

[0148] S6D1: Establish a multi-physical field coupling model (such as the physical relationship between temperature, pressure, and gas concentration);

[0149] S6D2: During data synchronization, use the coupling model to predict the data change trends of each sensor;

[0150] S6D3: According to the prediction results, dynamically adjust the time alignment method of sensor data;

[0151] S6D4: During the data integration process, map multi-modal data into a unified physical field model to generate comprehensive data.

[0152] Through the above steps, the present invention utilizes the physical relationships between sensor data to improve the accuracy of synchronization and integration, and is applicable to data acquisition in complex multi-physical field environments.

[0153] Data Synchronization and Integration Based on Blockchain

[0154] S6E1: Upload the data collected by each sensor to the blockchain network in real time;

[0155] S6E2: Utilize the consensus mechanism of the blockchain (such as PoW or PoS) to ensure the time consistency and immutability of the data;

[0156] S6E3: In the blockchain network, perform smart contract-driven synchronization and integration on multi-modal data;

[0157] S6E4: Store the integrated data in the blockchain; for subsequent query and verification.

[0158] Through the above steps, the present invention improves the transparency and credibility of the data, and is applicable to application scenarios with extremely high requirements for data reliability.

[0159] Real-Time Synchronization Based on Adaptive Filtering

[0160] S6F1: Deploy an adaptive filtering algorithm (such as Kalman filtering or particle filtering) in the edge computing unit;

[0161] S6F2: Receive the data of each sensor in real time, and perform time alignment and data correction according to the filtering algorithm;

[0162] S6F3: During the data integration process, dynamically adjust the filtering parameters to adapt to the changes in sensor data;

[0163] S6F4: Output the synchronized and integrated data for real-time monitoring and analysis;

[0164] Through the above steps, the present invention has strong real-time performance, is applicable to dynamic environments, and can effectively eliminate the influence of noise and delay on data synchronization.

[0165] Heterogeneous Data Integration Based on Graph Model

[0166] S6G1: Model multi-modal sensor data as a graph structure, where nodes represent sensors and edges represent the relationships between data;

[0167] S6G2: Use a graph neural network (GNN) to perform feature extraction and fusion on heterogeneous data;

[0168] S6G3: Dynamically adjust the data synchronization and integration strategies according to the analysis results of the graph model;

[0169] S6G4: Output the integrated data for further analysis and decision-making.

[0170] Through the above steps, the present invention can process complex heterogeneous data, improve the integration effect, and is applicable to the comprehensive analysis of multi-modal and multi-source data.

[0171] In summary, compared with the prior art, the present invention combines technologies such as edge computing, event triggering, machine learning, multi-physical field coupling, blockchain, adaptive filtering, and graph model, and can significantly improve the synchronization and integration effect of multi-modal sensor data. These methods are not only applicable to the automotive exhaust monitoring system, but also can be extended to other fields such as industrial monitoring, Internet of Things, and intelligent transportation.

[0172] S7: Transmit the collected exhaust parameter data to the edge computing unit. In the present invention, the specific transmission method can transmit data through wired (such as CAN bus) or wireless (such as Bluetooth, Wi-Fi) communication methods to ensure the real-time and reliability of data transmission, and avoid data loss or delay.

[0173] Furthermore, during the process of collecting exhaust parameters, it also includes monitoring data quality, detecting whether the sensors are working properly, and avoiding data anomalies caused by sensor failures; marking or eliminating abnormal data. Ensure the accuracy and reliability of the data. At the same time, regularly calibrate and maintain the multi-modal sensor array: check the sensitivity and accuracy of the sensors, and recalibrate if necessary; clean the surface of the sensors to avoid measurement errors caused by carbon deposition or contamination. Through the above steps, the multi-modal sensor array can accurately and reliably collect various parameters of automotive exhaust, providing high-quality data support for subsequent edge computing and cloud analysis.

[0174] Among them, marking abnormal data includes abnormal data detection and abnormal data marking;

[0175] Abnormal data detection

[0176] a1) Set a numerical range for each sensor, and check whether the sensor data exceeds the range according to the set numerical range;

[0177] If it is found that the sensor data exceeds the range, trigger the alarm mechanism; such as emitting an audible and visual alarm signal to remind relevant personnel to pay attention to the abnormal situation. At the same time, record and store the abnormal data; it is convenient to analyze the reasons for the problem subsequently, such as whether it is a sensor itself failure, external environmental interference, or a special situation of the monitored object, etc.

[0178] If the sensor data is within the normal range, continue with normal monitoring and continuously acquire and analyze the sensor data at the set time intervals or conditions. At the same time, use the normal data for trend analysis, performance evaluation, etc. For example, optimize the system operation parameters or predict possible future situations through the analysis of long-term normal data.

[0179] a2) Use a sliding window or differential method to detect sudden changes in the data while setting a mutation threshold.

[0180] a3) Check whether the physical relationships between multi-modal sensor data are reasonable.

[0181] If the physical relationships are reasonable, integrate the multi-modal data into a more accurate, comprehensive, and consistent final result through data fusion techniques (such as weighted average, Kalman filtering, Bayesian inference). For example, combining IMU (Inertial Measurement Unit) and GPS data can improve positioning accuracy. Extract meaningful features from the multi-modal data for subsequent machine learning or deep learning model training. For example, extract motion features from accelerometer and gyroscope data. Use the multi-modal data to train a prediction model to improve the performance and robustness of the model. For example, train an object tracking model by combining vision and IMU data. In practical applications, monitor the physical relationships of the multi-modal data in real time to ensure the stable operation of the system.

[0182] If the physical relationships are unreasonable, there may be sensor failures or calibration errors. Conduct individual tests or re-calibrate the suspicious sensors. Check whether there are environmental interferences (such as electromagnetic interference, temperature changes) causing data anomalies. If it is an environmental interference, shield the interference source or adjust the sensor position. Clean or interpolate the abnormal data to avoid its impact on subsequent analysis. For example, use the mean, median, or interpolation method to fill in missing or abnormal values.

[0183] a4) Check whether the sensor signal is stable (such as signal loss, excessive noise). Use the sensor self-diagnosis function (such as fault codes) to detect hardware failures. If the sensor signal suddenly interrupts, mark it as a fault.

[0184] Abnormal data marking

[0185] b1) Add a field (such as "status") to each data record to mark whether the data is abnormal.

[0186] B2) Mark the abnormal type and degree, and record the time of abnormal occurrence. The abnormal types include out of range, mutation, inconsistency, and sensor failure. The abnormal degrees include minor, medium, and severe. Recording the time of abnormal occurrence is for subsequent analysis.

[0187] The abnormal data marking also includes removing the abnormal data, which specifically includes temporary removal, data repair and permanent removal.

[0188] Preferably, the calibrating the multimodal sensor in S4 specifically includes:

[0189] S4B1: Collect data from multimodal sensors and reference equipment in a dynamic exhaust environment and compare and analyze errors;

[0190] The S4B1 also includes: data acquisition, data preprocessing and error analysis;

[0191] The data acquisition includes building a unified synchronization trigger system for multimodal sensors and reference devices; the system is based on a high-precision clock source (such as GPS timing or high-precision atomic clock synchronization signals) to ensure that in a dynamic exhaust environment, multimodal sensors and reference devices can start data acquisition at the same time. This synchronization method can effectively avoid acquisition time deviations caused by device clock drift, providing a basis for subsequent accurate comparative analysis. While collecting data, a special environmental monitoring module is used to record key parameters in the exhaust environment in real time (such parameters include temperature, pressure, humidity, airflow velocity, etc.), so that the complete recording of environmental information helps to more comprehensively analyze the source of errors;

[0192] The data preprocessing includes preprocessing the collected data according to the environmental parameters recorded in real time. For example, in areas where the temperature changes drastically, a filtering strategy based on temperature compensation is used to correct the sensor data and remove noise interference caused by environmental factors. Improve the accuracy and reliability of the data. And for the various types of data provided by multimodal sensors (such as optical signals, electrical signals, etc.), a special feature extraction algorithm is used to mine the key feature information of each modal data, and then the time series of data of different modalities are aligned according to these feature information, so that when comparing and analyzing, the multimodal data and the reference device data are accurately matched in the time dimension; effectively solve the problem of time synchronization in multimodal data processing in traditional methods.

[0193] The error analysis includes introducing more dimensional error measurement indicators in addition to traditional indicators such as the mean square error (MSE) and mean absolute error (MAE). For example, considering the rate of change error of data, by calculating the difference in the rate of change between multi-modal sensor data and reference device data at adjacent time points, the synchronization and accuracy of the two during the dynamic change process are evaluated; introducing the standard deviation of the relative error to measure the dispersion of the error at different measurement times, and more comprehensively reflecting the characteristics of the error. Using machine learning algorithms (such as convolutional neural network CNN or recurrent neural network RNN in deep learning) to train the collected data to establish an error pattern recognition model; this model can automatically learn the characteristic patterns of errors under different working conditions from a large amount of data and identify the main factors causing the errors (such as changes in the concentration of specific exhaust components, equipment failures, etc.). In a dynamic exhaust environment, if the error propagates and accumulates in the data sequence over time, the method of dynamic system modeling is adopted to analyze the propagation law of the error between different times and different measurement points, and by establishing an error propagation model, the development trend of the error is predicted. And corresponding compensation measures are taken to effectively reduce the impact of the error on the measurement results.

[0194] S4B2: Use redundant sensor data fusion to generate a reference value and calibrate the output of the multi-modal sensor;

[0195] S4B3: Use historical data to train the calibration model and adjust the sensor output;

[0196] S4B4: Use the physical relationship between multi-modal sensors for cross-calibration; ensure data consistency.

[0197] S4B5) During the operation of the vehicle, analyze the sensor data and dynamically adjust the calibration parameters.

[0198] Before the above-mentioned S4B1, the following steps are also included:

[0199] Before the formal calibration (S4B1), some preparatory work needs to be completed to ensure the smooth progress of the calibration process.

[0200] S4A1: Identify the parameters to be calibrated (such as gas concentration, temperature, pressure, flow rate, etc.) and their accuracy requirements;

[0201] S4A2: Prepare reference sensors or instruments (such as laboratory-grade gas analyzers, thermometers, pressure gauges, etc.) as the calibration benchmark. Among them, high-precision reference sensors or instruments are selected.

[0202] S4A3: Build a dynamic exhaust environment simulation device in the laboratory or in an actual vehicle, which can simulate the exhaust conditions during vehicle operation;

[0203] S4A4: Install the multimodal sensor array in the exhaust pipe; ensure that it can accurately collect exhaust parameters;

[0204] S4A5: Configure a data acquisition system for synchronously collecting data from the multimodal sensors and reference devices.

[0205] Preferably, after S4B5, some optimization and verification work needs to be carried out to ensure the long-term stability of the calibration effect. Specifically,

[0206] S4C1: Dynamically adjust the calibration parameters of the sensors according to changes in environmental parameters; improve robustness;

[0207] S4C2: Record the calibration data in the blockchain; ensure the immutability and traceability of the data.

[0208] S4C3: Verify the calibration effect in the actual operating environment; ensure the accuracy and reliability of the sensor data.

[0209] S4C4: Regularly check the calibration status of the sensors and recalibrate or update the calibration model if necessary.

[0210] In the present invention

[0211] The above-mentioned S4A1 - S4A5 are the basis for calibration, ensuring the reliability of the calibration environment and equipment;

[0212] The above-mentioned S4B1 - S4B5 are the core of calibration, improving the measurement accuracy of the sensors through various methods.

[0213] The above-mentioned S4C1 - S4C4 are the extension of calibration, ensuring the long-term stability and traceability of the calibration effect.

[0214] The preferred specific embodiments of the present invention have been described in detail above. It should be understood that those of ordinary skill in the art can make many modifications and variations based on the concept of the present invention without creative efforts. Therefore, all technical solutions that can be obtained by those skilled in the art in the technical field of the present invention based on the concept of the present invention through logical analysis, reasoning or limited experiments on the basis of the prior art shall fall within the protection scope determined by the claims.

Claims

1. A real-time monitoring system for automobile exhaust flow based on intelligent sensors, characterized by: include A multimodal sensor array connected to an edge computing unit to collect vehicle exhaust parameters; An edge computing unit, connected to the cloud monitoring platform, is used to process the parameters collected by the multimodal sensor array; The cloud-based monitoring platform establishes a multi-physics field coupling model to achieve emission trend prediction and abnormal alarm.

2. The real-time monitoring system for automobile exhaust flow based on intelligent sensors as claimed in claim 1 is characterized by: The specific steps of the multimodal sensor array to collect the parameters of automobile exhaust include: S1: Determine the automobile exhaust parameters that need to be collected; S2: Selecting a multimodal sensor according to the automobile exhaust parameters to be collected; S3: placing the sensor array selected in S2 at the exhaust pipe of the automobile; S4: calibrate the multimodal sensor; S5: starting the multi-modal sensor array to collect exhaust parameters; S6: Synchronize and integrate the data collected by multimodal sensors; S7: Transmit the collected exhaust parameter data to the edge computing unit.

3. The real-time monitoring system for automobile exhaust flow based on intelligent sensors as claimed in claim 2 is characterized by: The process of collecting exhaust parameters also includes monitoring data quality, detecting whether the sensor is working properly, and marking or eliminating abnormal data.

4. The real-time monitoring system for automobile exhaust flow based on intelligent sensors as claimed in claim 3 is characterized in that: The calibrating of the multimodal sensor in S4 specifically includes: S4B1: Collect data from multimodal sensors and reference equipment in a dynamic exhaust environment and compare and analyze errors; S4B2: Generate reference values ​​using redundant sensor data fusion to calibrate the outputs of multimodal sensors; S4B3: Use historical data to train a calibration model and adjust sensor output; S4B4: Exploiting the physical relationship between multimodal sensors for cross-calibration; S4B5) Analyze sensor data and dynamically adjust calibration parameters during vehicle operation.

5. The automobile exhaust flow monitoring system based on intelligent sensor as claimed in claim 4 is characterized in that: The S4B1 also includes the following steps: S4A1: Identify the parameters that need to be calibrated and their accuracy requirements; S4A2: Prepare reference sensors or instruments to serve as calibration benchmarks; S4A3: Build a dynamic exhaust environment simulation device in the laboratory or in an actual vehicle to simulate the exhaust conditions when the vehicle is running; S4A4: Installing a multimodal sensor array in an exhaust duct; S4A5: Configure a data acquisition system for synchronously acquiring data from multimodal sensors and reference devices.

6. The real-time monitoring system for automobile exhaust flow based on intelligent sensors as claimed in claim 5 is characterized in that: The S4B5 also includes: S4C1: Dynamically adjust sensor calibration parameters according to changes in environmental parameters; S4C2: Record calibration data in blockchain; S4C3: Verify the calibration effect in the actual operating environment; S4C4: Regularly check the calibration status of the sensor and recalibrate or update the calibration model if necessary.

7. The real-time monitoring system for automobile exhaust flow based on intelligent sensors as claimed in claim 6 is characterized in that: The step of synchronizing and integrating the data collected by the multimodal sensors in S6) includes: Distributed time synchronization based on edge computing S6A1: Deployment of distributed time synchronization algorithms in edge computing units; S6A2: Assign a unique timestamp to each sensor node and calibrate the time uniformly through the edge computing unit; S6A3: During the data collection process, the edge computing unit receives the data from each sensor in real time and aligns them according to the timestamp; S6A4: Use interpolation or extrapolation methods to fill in missing data caused by communication delays or different sampling rates; Event-triggered data synchronization S6B1: Define several trigger events; S6B2: When a sensor detects a trigger event, it sends a synchronization signal to the edge computing unit; S6B3: After receiving the synchronization signal, the edge computing unit collects data from other sensors and performs time alignment; S6B4: Integrate the synchronized data to generate a unified data packet; Data-driven intelligent synchronization S6C1: Analyzing time series features of multimodal sensor data using machine learning algorithms; S6C2: Identify time offsets based on data features and perform dynamic alignment; S6C3: Perform data correction during data integration; S6C4: Store the synchronized and integrated data into a unified data structure; Synchronization and integration based on multi-physics coupling S6D1: Establish multi-physics coupling model; S6D2: During data synchronization, the coupling model is used to predict the data change trend of each sensor; S6D3: Dynamically adjust the time alignment of sensor data based on the prediction results; S6D4: In the data integration process, multimodal data are mapped into a unified physical field model to generate comprehensive data; Data synchronization and integration based on blockchain S6E1: Upload the data collected by each sensor to the blockchain network in real time; S6E2: Use the consensus mechanism of blockchain to ensure the time consistency and immutability of data; S6E3: Smart contract driven synchronization and integration of multimodal data in blockchain networks; S6E4: Store the integrated data into the blockchain; Real-time synchronization based on adaptive filtering S6F1: Deploy adaptive filtering algorithms in edge computing units; S6F2: Receives data from each sensor in real time and performs time alignment and data correction based on the filtering algorithm; S6F3: During the data integration process, filter parameters are dynamically adjusted to adapt to changes in sensor data; S6F4: Output synchronized and integrated data for real-time monitoring and analysis; Heterogeneous data integration based on graph model S6G1: Model multimodal sensor data as a graph structure where nodes represent sensors and edges represent relationships between data; S6G2: Feature extraction and fusion of heterogeneous data using graph neural networks; S6G3: Dynamically adjust data synchronization and integration strategies based on the analysis results of the graph model; S6G4: Output integrated data for analysis and decision making.

8. The real-time monitoring system for automobile exhaust flow based on intelligent sensors as claimed in claim 7 is characterized in that: The marking of abnormal data includes abnormal data detection and abnormal data marking; Abnormal data detection a1) Set a value range for each sensor and check whether the sensor data is within the set value range; If the sensor data is found to be out of range, the alarm mechanism will be triggered; at the same time, the abnormal data will be recorded and stored; If the sensor data is within the range, normal monitoring continues, and sensor data is continuously acquired and analyzed; a2) Use sliding windows or difference methods to detect sudden changes in data and set mutation thresholds; a3) Check whether the physical relationship between multimodal sensor data is reasonable; If the physical relationship is reasonable, the multimodal data is integrated into the final result through data fusion technology; features are extracted from the multimodal data for subsequent machine learning or deep learning model training, for example, motion features are extracted from accelerometer and gyroscope data; Use multimodal data to train predictive models; monitor physical relationships in multimodal data; If the physical relationship is unreasonable, there is a sensor failure or calibration error. Test or recalibrate the suspected sensor separately. Check whether there is environmental interference that causes data anomalies. If it is environmental interference, shield the interference source or adjust the sensor position. Clean or interpolate the abnormal data. a4) Check whether the sensor signal is stable; use the sensor self-diagnosis function to detect hardware failures. If the sensor signal is suddenly interrupted, it will be marked as a failure; Abnormal data marking b1) Add a field to each data record to mark whether the data is abnormal. b2) Mark the type and degree of abnormality and record the time when the abnormality occurred.

9. The real-time monitoring system for automobile exhaust flow based on intelligent sensors as claimed in claim 8 is characterized in that: The abnormal data marking also includes removing the abnormal data, which specifically includes temporary removal, data repair and permanent removal.

10. The real-time monitoring system for automobile exhaust flow based on intelligent sensors as claimed in claim 9, characterized in that: In the S4B1, data from the multimodal sensor and the reference device are collected in a dynamic exhaust environment, and the comparison and analysis of errors include: data collection, data preprocessing and error analysis; The data acquisition includes building a unified synchronous trigger system for the multimodal sensor and the reference device; and while collecting data, the parameters in the exhaust environment are recorded through the environmental monitoring module, so that the complete recording of environmental information is helpful for a more comprehensive analysis of the error source; The data preprocessing includes preprocessing the collected data according to the recorded environmental parameters. And for the type data provided by the multimodal sensor, the characteristic information of each modal data is mined, and then the time series of the data of different modalities are aligned according to the characteristic information, so that when the multimodal data is compared and analyzed, the multimodal data and the reference device data are accurately matched in the time dimension; The error analysis includes using a machine learning algorithm to train the collected data and establish an error pattern recognition model; in a dynamic exhaust environment, if the error propagates and accumulates in the data sequence over time, a dynamic system modeling method is used to analyze the propagation law of the error at different times and different measurement points, and by establishing an error propagation model, the development trend of the error is predicted.

Images