Leakage detection system of automobile thermal management system and use method of leakage detection system
By analyzing sensor data using convolutional neural networks and random forest models, the efficiency and accuracy issues of refrigerant leak detection in automotive thermal management systems were resolved. This enabled rapid and accurate location of leak points, improving system safety and reducing costs.
Patent Information
- Application Number
- CN202511003615.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-21
- Publication Date
- 2025-11-21
AI Technical Summary
Existing automotive thermal management systems rely on manual inspections for refrigerant leak detection, which suffers from low detection efficiency and insufficient accuracy. Excessive sensor deployment or complex data processing makes it difficult to quickly and accurately locate leak points, posing safety hazards.
By employing convolutional neural networks (CNN) and random forest (RF) models to analyze sensor data, and through preprocessing, feature extraction, and comparison, we can achieve rapid determination of refrigerant leaks and accurate location of leak points, thereby reducing sensor deployment and improving detection efficiency and accuracy.
It enables rapid response and accurate location of refrigerant leaks, improves the safety performance and detection efficiency of automotive thermal management systems, and reduces equipment costs.
Smart Images
Figure CN120992111A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to a leakage detection system of an automobile thermal management system and a method of using the same. BACKGROUND
[0002] With the rapid development of the automobile industry, the complexity and reliability requirements of automobile thermal management systems are increasing. However, the leakage of refrigerant in the thermal management system is one of the common faults, which not only affects the normal operation of the air conditioning system, but also may pose a potential threat to the overall performance and safety of the vehicle (such as R134a is toxic, and R290 is flammable when exposed to open flames). Traditional detection methods often rely on manual inspection and experience-based judgment, which has low detection efficiency and insufficient accuracy.
[0003] Nowadays, sensors and intelligent control as the basis of intelligent vehicles are widely used, which can sense real-time parameters at various positions in the vehicle through widely distributed sensors in the vehicle. If a concentration sensor is used, sensors need to be arranged at multiple positions in the vehicle, which not only increases equipment costs, but also easily causes long-term leakage without sensing due to the uncertainty of the leakage point. If a conventional sensor is used, the control center will receive a lot of data feedback in real time, which needs to be processed in time. It is difficult to achieve accurate command issuance and judgment by relying on traditional calculation models only, which also causes the inability to accurately locate the leakage point and causes safety accidents of leaked gas due to the inability to process in time.
[0004] Therefore, how to quickly calculate whether the refrigerant leaks and the leakage point through the sensing data of the traditional sensor without additional equipment assembly, so as to improve the overall reaction speed and sensitivity of the vehicle, can effectively improve the safety performance of the vehicle thermal management system and reduce the cost of the vehicle-mounted system. SUMMARY
[0005] The technical problem to be solved by the present application is to provide a leakage detection system of an automobile thermal management system and a method of using the same, which utilizes convolutional neural network (CNN) and random forest (RF) model data analysis to analyze the space of the leakage position in time, and trains the known data model to predict the future scene, and predicts and tracks the leakage safety after verification, thereby improving the detection efficiency.
[0006] In order to solve the above technical problems, the technical scheme of the present application is: a leakage detection system of an automobile thermal management system, comprising a collection module, a preprocessing module, an extraction module and a detection module; the collection module is used for collecting sensor data of an automobile air conditioning pipeline and connecting various sensors in the automobile internal thermal management system; the preprocessing module performs cleaning, denoising and normalization processing on the collected data, calculates the refrigerant charge, and then is used for comparison and judgment of refrigerant leakage; the extraction module is used for feature extraction of the preprocessed data and extraction of calculation data in real-time detection, simplifying the data calculation process of leakage detection; the detection module is used for comparing the real-time detection data with the data in the extraction module to judge whether there is leakage, and judging the leakage point according to the output position of the data; The collection module records the pressure value, temperature value or concentration value of each pressure sensor at consecutive time points, and arranges temperature measuring points at the inlet and outlet of each component in the vehicle and along the heat exchanger to obtain the required temperature signal. The preprocessing module calculates the refrigerant charge in the vehicle-mounted air conditioning system through a preprocessing model, so as to make it as a standard comparison value for refrigerant leakage. The extraction module is used for extracting the refrigerant charge in the preprocessing module and extracting the calculation results of the data of each sensor in real-time detection.
[0007] Further, the calculation formula of the preprocessing module is the sum of the refrigerant mass in each component, which is equal to the sum of the refrigerant mass in the compressor, the refrigerant mass in the condenser, the refrigerant mass in the cooler and the refrigerant mass in each pipeline; the measuring point temperatures of each component monitored by the preprocessing model are calculated and processed to obtain the refrigerant mass in each component.
[0008] Further, the preprocessing module further comprises a prediction model, the prediction model adopts a convolutional neural network (CNN) and a random forest (RF) model for data analysis, analyzes the real-time experimental sampling data, normalizes the experimental data of the sensor first, generates a gray matrix graph using the collected two-second data, then calculates the data features through the convolutional neural network (CNN) model, outputs, then calculates and decides the output through the random forest (RF) model, and finally outputs the result.
[0009] Further, the prediction model comprises a spatial data module and a time analysis module, normalizes the data, extracts local features through the convolutional neural network (CNN) model, and obtains spatial position sequence data and leakage time sequence data.
[0010] Further, the spatial position sequence data is integrated into a training set, and learning training is performed by using the training set.
[0011] Further, the leakage time sequence data is integrated into a test set, and verification accuracy is performed by experiment.
[0012] Further, the spatial position sequence data and the leakage time sequence data are integrated into a random forest (RF) model, and reconstruction, normalization and inverse change are performed, and finally a prediction value is obtained.
[0013] Further, the prediction value is compared with the actual sensor data value, so as to obtain an accurate leakage point and time.
[0014] Further, the use method of the leakage detection system is as follows: Firstly, the leakage detection system is loaded into the vehicle-mounted air conditioning system, the thermal management system is evacuated and the rated mass of refrigerant is filled, the pre-treatment temperature is set by using the air conditioning system, the air conditioning system is pre-treated at the set temperature, after pre-treatment, the vehicle-mounted environment is adjusted, and the leakage experiment is performed; Secondly, the actual signal detected in the leakage detection system is compared with the set threshold value by calculating the real-time mass of the refrigerant by using the pre-treatment module and comparing the ratio with the filling amount, so as to determine whether leakage occurs; if the ratio is less than the set threshold value, it is determined that no leakage occurs; if the ratio is greater than the set threshold value, it is determined that leakage occurs. Thirdly, when refrigerant leakage is detected, the data of the sensors in each component within two seconds is collected, the data is collected and normalized by using a convolutional neural network (CNN) model, local features are extracted, so as to obtain spatial position sequence data and leakage time sequence data, and the two data sets are integrated into a training set and a test set, and learning and verification are performed by using a random forest (RF) model. Fourthly, a prediction model is obtained according to different actual data of multiple experiments, and the leakage point and time are determined according to the prediction model. Fifthly, after learning and verification are completed, the leakage detection system is actually loaded into the vehicle-mounted air conditioning system, and whether refrigerant leakage occurs and the leakage point and time are determined according to the feedback data of different sensors.
[0015] Further, the data in the training set is used for learning and training of a random forest (RF) model, the data in the test set is used for verification accuracy of the random forest (RF) model, so as to realize deep learning of the prediction model.
[0016] Compared with the prior art, the leakage detection system of the automobile thermal management system and the use method thereof provided by the present application utilize sensors to monitor the temperature, pressure or refrigerant concentration data in the vehicle in real time, and then normalize the experimental data through a prediction model, and the prediction data obtained in the prediction model is applied in real-time detection, so that the leakage judgment and the confirmation of the leakage point and time can be quickly responded, the accuracy of actual application is improved, the arrangement of sensors is reduced, and the detection efficiency and accuracy are improved. BRIEF DESCRIPTION OF DRAWINGS
[0017] Figure 1 The schematic diagram of the air conditioning system of the present application is shown.
[0018] Figure 2 The refrigerant leakage detection analysis diagram of the present application is shown.
[0019] Figure 3 The detection flowchart of the prediction model of the present application is shown.
[0020] Figure 4 The structure diagram of the prediction model of the present application is shown.
[0021] Figure 5 The leakage detection flowchart of the present application is shown.
[0022] Among them: 1. Acquisition module, 2. Preprocessing module, 3. Extraction module, 4. Detection module, 5. Preprocessing model, 6. Prediction model. DETAILED DESCRIPTION
[0023] As Figure 5 shown, a leakage detection system of an automobile thermal management system comprises an acquisition module 1, a preprocessing module 2, an extraction module 3 and a detection module 4; the acquisition module 1 is used to acquire sensor data of the automobile air conditioning pipeline, and is connected to each sensor in the automobile internal thermal management system (as shown in Figure 1 The preprocessing module 2 cleans, denoises and normalizes the acquired data, calculates the charge of the refrigerant, and then compares and judges whether the refrigerant leaks; the extraction module 3 is used to extract features from the preprocessed data, and extract calculation data in real-time detection, so as to simplify the data calculation process of leakage detection; the detection module 4 is used to compare the real-time detection data with the data in the extraction module 3, judge whether there is leakage, and judge the leakage point according to the output position of the data; The acquisition module 1 records the pressure value, temperature value or concentration value of each pressure sensor at consecutive time points, and arranges measuring points at the inlet and outlet of each component in the vehicle and along the heat exchanger to obtain the required data signal; The pre-processing module 2 calculates the refrigerant charge amount in the vehicle air conditioning system through the pre-processing model 5, so as to serve as a standard comparison value for refrigerant leakage; The extraction module 3 is used to extract the refrigerant charge amount in the pre-processing module 2, and extract the results calculated from the data of each sensor in real-time detection.
[0024] In one embodiment, as shown in Figure 2 The calculation formula of the pre-processing module 2 is the sum of the refrigerant mass in each component, which is equal to the sum of the refrigerant mass in the compressor, the refrigerant mass in the condenser, the refrigerant mass in the cooler, and the refrigerant mass in each pipeline; the measured point temperature of each component is processed and calculated through the pre-processing model 5 to obtain the refrigerant mass in each component. The formula is as follows:
[0025] Wherein: Σmi is the sum of the refrigerant mass in each component; mcharge is the refrigerant charge amount; mcomp is the refrigerant mass in the compressor; mLCC is the refrigerant mass in the condenser; mChiller is the refrigerant mass in the cooler; mpipej is the refrigerant mass in the jth pipeline; t is the current monitoring time; n is the total number of temperature measuring points; T is the temperature of the temperature measuring point.
[0026] In one embodiment, as shown in Figure 3 and 4 As shown, the pre-processing module 2 is further provided with a prediction model 6, which uses a convolutional neural network (CNN) and a random forest (RF) model for data analysis. The experimental data of the sensor are normalized, and a gray matrix graph is generated using two seconds of collected data. Then, the data features are calculated by the convolutional neural network (CNN) model, and the output is output. Then, the random forest (RF) model is used for calculation and decision output, and finally the output result is obtained.
[0027] The prediction model 6 includes a spatial data module and a time analysis module. The data are normalized, and the local features are extracted by the convolutional neural network (CNN) model to obtain spatial position sequence data and leakage time sequence data.
[0028] The spatial position sequence data are integrated into a training set for learning and training. The leakage time sequence data are integrated into a test set for verifying the accuracy. The spatial position sequence data and the leakage time sequence data are integrated into the random forest (RF) model for reconstruction, normalization and inverse change, and finally the prediction value is obtained. The prediction value is compared with the actual sensor data value to obtain the accurate leakage point and time.
[0029] In one embodiment, as shown in Figure 3 The method for using the leakage detection system is as follows: First, load the leakage detection system into the vehicle air conditioning system, vacuum the thermal management system and charge the rated mass of refrigerant, set the pre-treatment temperature using the air conditioning system, and pre-treat the air conditioning system at the set temperature. After pre-treatment, adjust the vehicle environment and perform a leakage experiment. Second, the actual signal detected by the leakage detection system is calculated by the pre-treatment module 2 to obtain the real-time mass of the refrigerant, and the ratio of the charge is compared to determine whether there is leakage. If the ratio is less than the set threshold, it is determined that there is no leakage. If the ratio is greater than the set threshold, it is determined that there is leakage. Third, when refrigerant leakage is detected, collect the data of the sensors in each component within two seconds, collect the data and normalize it through a convolutional neural network (CNN) model, extract local features, and obtain spatial position sequence data and leakage time sequence data. Then, the two data sets are organized into training and test sets, and the random forest (RF) model is used for learning and verification. Fourth, according to the different actual data of multiple experiments, a prediction model 6 is obtained, and the leakage point and time are determined according to the prediction model 6. Fifth, after the leakage detection system completes learning and verification, it is actually loaded into the vehicle air conditioning system, and according to the return data of different sensors, it is determined whether there is refrigerant leakage and the leakage point and time.
[0030] The data in the training set is used for learning and training of the random forest (RF) model, and the data in the test set is used for verification accuracy of the random forest (RF) model, so as to realize deep learning of the prediction model 6.
[0031] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting technical solutions. Those skilled in the art should understand that those who modify or equivalently replace the technical solutions of the present application without departing from the spirit and scope of the present application should be covered by the claims of the present application.
Claims
1. A leak detection system for an automotive thermal management system, characterized by, Including collection module, pretreatment module, extraction module and detection module;The collection module is used for collecting sensor data of automobile air conditioning pipeline, connecting each sensor in the internal thermal management system of the automobile;The pretreatment module cleans, denoises and normalizes the collected data, calculates the charge of the refrigerant, and then is used for comparison to determine whether the refrigerant leaks;The extraction module is used for feature extraction of the pretreated data, and simultaneously extracts the calculation data in real-time detection, simplifying the data calculation process of leakage detection;The detection module is used for comparing the real-time detection data with the data in the extraction module to determine whether there is leakage, and determining the leakage point according to the output position of the data; The collection module records the pressure value, temperature value or concentration value of each pressure sensor at consecutive time points, and arranges temperature measuring points at the inlet and outlet of each component in the vehicle and along the heat exchanger to obtain the required temperature signal; The pretreatment module calculates the refrigerant charge in the vehicle-mounted air conditioning system through a pretreatment model, so as to serve as a standard comparison value for refrigerant leakage; The extraction module is used for extracting the refrigerant charge in the pretreatment module, and simultaneously extracting the calculation results of the data of each sensor in real-time detection.
2. The leak detection system of a vehicle thermal management system according to claim 1, wherein The calculation formula of the pretreatment module is the sum of the refrigerant mass in each component, which is equal to the sum of the refrigerant mass in the compressor, the refrigerant mass in the condenser, the refrigerant mass in the cooler and the refrigerant mass in each pipeline. The pretreatment model processes and calculates the measured point temperatures monitored by each component to obtain the refrigerant mass in each component.
3. The leak detection system of a vehicle thermal management system of claim 1, wherein, The pretreatment module further comprises a prediction model, which uses a convolutional neural network (CNN) and a random forest (RF) model for data analysis. The prediction model analyzes the real-time experimental sampling data, normalizes the experimental data of the sensor first, generates a gray matrix graph using the collected two-second data, then calculates the data features through the convolutional neural network (CNN) model, outputs the data features, and then calculates and outputs the data features through the random forest (RF) model. Finally, the output result is obtained.
4. The leak detection system of claim 3, wherein, The prediction model comprises a spatial data module and a time analysis module, which normalize the data, extract local features through the convolutional neural network (CNN) model, and obtain spatial position sequence data and leakage time sequence data.
5. The leak detection system of a vehicle thermal management system of claim 4, wherein, The spatial position sequence data is integrated into a training set for learning and training.
6. The leak detection system of claim 4, wherein, The leakage time sequence data is integrated into a test set for verifying the accuracy through experiments.
7. The leak detection system of claim 4, wherein, The spatial position sequence data and the leakage time sequence data are integrated into the random forest (RF) model for reconstruction, normalization and inverse change, and finally the prediction value is obtained.
8. The leak detection system of a vehicle thermal management system of claim 7, wherein, The prediction value is compared with the actual sensor data value to obtain the accurate leakage point and time.
9. The leak detection system of a vehicle thermal management system according to any one of claims 1-8, wherein, The use method of the leakage detection system is as follows: The first step is to load the leakage detection system into the vehicle air conditioning system, vacuum the thermal management system and charge the rated mass of refrigerant, set the pre-treatment temperature of the air conditioning system, and pre-treat the air conditioning system at the set temperature. After pre-treatment, adjust the vehicle environment and start the leakage experiment; The second step is to compare the real-time refrigerant mass calculated by the pre-treatment module with the charging amount according to the fact signal detected by the leakage detection system to determine whether there is leakage. If the ratio is less than the set threshold, it is determined that there is no leakage. If the ratio is greater than the set threshold, it is determined that there is leakage; The third step is to collect the data of the sensors in each component within two seconds when refrigerant leakage is detected, collect the data and normalize it through a convolutional neural network (CNN) model, extract local features, and obtain spatial position sequence data and leakage time sequence data. Then, the two data sets are organized into training and test sets, and the random forest (RF) model is used for learning and verification; The fourth step is to establish a prediction model based on different actual data from multiple experiments, and to determine the leakage point and time based on the prediction model; The fifth step is to load the leakage detection system into the vehicle air conditioning system after learning and verification, and to determine whether there is refrigerant leakage and the leakage point and time based on the data returned by different sensors.
10. The method of claim 9, wherein, The data in the training set is used for learning and training of the random forest (RF) model, and the data in the test set is used for verification accuracy of the random forest (RF) model, thereby realizing deep learning of the prediction model.