Flying car intelligent heat dissipation management method based on multi-modal data

Abstract

The invention relates to the technical field of intelligent heat dissipation of oil coolers, in particular to an intelligent heat dissipation management method for a hovercar based on multi-modal data, and the method achieves the prediction of the heat dissipation demands of the hovercar in different working modes through integrating the data collected by a plurality of sensors and combining a deep learning model. The intelligent heat dissipation strategy adapting to the current working condition can be automatically generated and implemented.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent heat dissipation of oil coolers, and in particular to an intelligent heat dissipation management method for a flying car based on multimodal data. Background Art

[0002] With the growth of the global population and economic development, traffic pressure is increasing. Traditional automobile traffic faces problems such as road congestion and environmental pollution, which seriously restricts economic development and the improvement of quality of life. As a concept generated by coping with traffic difficulties and pursuing sustainable development, flying cars are expected to combine the advantages of cars and airplanes, get rid of ground traffic congestion, reduce environmental pollution, improve transportation efficiency and capacity, and make the transportation network more flexible and efficient. Compared with ordinary cars, flying cars have more advantages in hovering, climbing and crossing obstacles, which means that they can use vertical space more effectively and provide more diverse travel methods. They will make full use of the vertical space and air space within the city, and provide more choices for urban residents in terms of transportation, speed, comfort and efficiency. At the same time, flying cars can also be used for long-distance inter-city travel, which is expected to reduce carbon emissions and horizontal inefficiency, and provide people with more intelligent and sustainable travel solutions.

[0003] However, like other aircraft, flying cars also face key technical challenges during flight. For example, compared with ground vehicles, flying cars need to overcome more energy loss in the engine to maintain its normal operation during flight, a large part of which is heat energy. The normal operation of the flying car engine and body requires a suitable temperature range. If the temperature is too high, it will cause the performance of various components to decline or even be damaged, thus affecting the overall performance of the flying car. Therefore, the heat dissipation management issue is the focus of attention in the design and use of flying multi-functional aircraft.

[0004] The heat dissipation management of current flying cars usually relies on traditional cooling systems, which mainly include air cooling, water cooling or liquid cooling technologies. Although these technologies are effective, they have certain limitations when facing the complex and changing flight environment and the highly integrated characteristics of flying cars. For example, traditional cooling systems often require a fixed heat dissipation mode to be preset, and cannot be flexibly adjusted according to the real-time working conditions of flying cars. Specifically, there are the following defects: Fixed cooling mode: Existing cooling management systems usually adopt a fixed cooling strategy and cannot be adjusted in real time according to the operating mode of the flying car (such as hovering, cruising or rapid climbing).

[0005] Insufficient data utilization: Existing systems rarely use multimodal data (such as temperature, air pressure, humidity, flight attitude, etc.), which limits the system's understanding and adaptability to the actual working conditions of flying cars.

[0006] Low intelligence: Existing heat dissipation management systems usually lack the ability to learn and optimize autonomously, and it is difficult to automatically adjust the heat dissipation strategy according to environmental changes and usage habits.

[0007] Therefore, there is a need for an intelligent heat dissipation management method for flying cars based on multimodal data that can solve the above problems. Summary of the invention

[0008] The present invention provides a flying car intelligent heat dissipation management method based on multimodal data. The present invention integrates data collected by multiple sensors and combines a deep learning model to predict the heat dissipation requirements of the flying car in different working modes, and can automatically generate and implement an intelligent heat dissipation strategy that adapts to the current working conditions.

[0009] The technical solution adopted by the present invention to solve the above technical problems is: A flying car intelligent heat dissipation management method based on multimodal data comprises the following steps: a. Collecting real-time monitoring data through multi-modal sensors integrated on the flying car, the multi-modal sensors include temperature sensors, pressure sensors, humidity sensors and accelerometers; b. Use data fusion algorithm to fuse data from different sensors; c. Build a deep learning model that combines convolutional neural networks with long short-term memory networks to learn the heat dissipation rules of flying cars in different working modes and predict the heat dissipation needs of flying cars; d. Transmit the fused data to a remote server or cloud platform for real-time analysis and processing; e. Generate an intelligent heat dissipation strategy for the current working mode of the flying car based on the prediction results of the deep learning model, wherein the intelligent heat dissipation strategy includes adjusting the fan speed and coolant flow rate; f. Implement intelligent heat dissipation strategies and dynamically adjust the heat dissipation strategies according to the actual operation of the flying car; g. Collect feedback data during the operation of the flying car, and use the feedback data to iteratively train and optimize the deep learning model.

[0010] Furthermore, the temperature sensor is installed on the engine surface, inside the battery pack and the electronic control unit of the flying car; the pressure sensor is installed in the hydraulic system of the flying car and on the wing surface and air inlet; the humidity sensor is installed at the environmental monitoring points inside and outside the flying car; the accelerometer is fixedly installed at the center of gravity of the vehicle to monitor the linear and angular acceleration during flight.

[0011] Furthermore, in step b, a data fusion algorithm is used to perform fusion processing on different sensor data, including: b1. Preprocessing the sensor data: performing outlier processing on the collected sensor data to remove outliers and missing values ​​in the sensor data; b2. Using a time series alignment method to match and align different sensor data to ensure the temporal consistency of the sensor data; b3. Standardization processing: ensuring that the dimensions and ranges of different sensor data are consistent; b4. Performing correlation analysis and weighted processing on sensor parameters: calculating the correlation between each sensor parameter, and assigning weights to comprehensively consider the impact of different parameters on the heat dissipation demand.

[0012] Furthermore, in step c, a deep learning model is constructed, and the following steps are performed: c1. Historical operation data of the flying car is collected, including sensor data and heat dissipation management operation records; c2. The historical operation data is divided into a training data set and a test data set; c3. A deep learning model is constructed based on a convolutional neural network and a long short-term memory network; c4. The training data is input into the deep learning model for training, and the output prediction value is calculated through forward propagation.

[0013] Further, in step c3, a deep learning model is constructed based on a convolutional neural network and a long short-term memory network, and the following steps are performed: c31. Construct an input layer: use multimodal data as the input layer, and define the input layer dimension of the network; c32. Construct a convolution layer: extract and abstract features of the input layer data through convolution operations and excitation functions; c33. Construct a pooling layer: downsample the extracted features to reduce network parameters; c34. Construct a fully connected layer: expand the feature vector output by the pooling layer into a one-dimensional vector and perform a linear transformation; c35. Construct a recurrent layer: use long short-term memory units and gate structures to process time series data, capture contextual information, and realize memory functions; c36. Construct an output layer: perform classification or regression operations based on the output of the recurrent layer to generate prediction results.

[0014] Further, in step c4, the training data is input into the deep learning model for training, and the output prediction value is calculated through forward propagation, and the following steps are performed: c41. The original multimodal data set is divided into multiple batches of data, and the batches of data are randomly selected as the input of the model; c42. A convolution operation is performed to extract local features, and the local features are added to obtain a hidden state Z; c43. The hidden state Z is nonlinearly transformed using an activation function to obtain a new hidden state Z′; c44. A pooling operation is performed to reduce the dimension of the hidden state Z′ by downsampling to obtain the pooling layer output result R; c45. A back propagation operation is performed to update the weights and biases according to the error between the pooling layer output result R and the true value, and c41 to c45 are repeated until the convergence of the model reaches the expected value.

[0015] Furthermore, in step d, the fused data is transmitted to a remote server or cloud platform for real-time analysis and processing, and the following steps are performed: d1. The multimodal sensor data is transmitted to a remote server or cloud platform via wired or wireless communication; d2. The remote server or cloud platform further analyzes and processes the data, and uses a machine learning algorithm or a deep learning algorithm to perform multimodal data fusion.

[0016] Furthermore, in step e, an intelligent heat dissipation strategy for the current working mode of the flying car is generated according to the prediction results of the deep learning model, and the following steps are performed: e1. corresponding thresholds are set for different types of sensor data; e2. fan speed and coolant flow adjustment rules are set; e3. fan speed and coolant flow of the flying car are dynamically adjusted according to the prediction results of the deep learning model.

[0017] Furthermore, in step f, an intelligent heat dissipation strategy is implemented, and the heat dissipation strategy is dynamically adjusted according to the actual operation of the flying car, and the following steps are performed: f1. Adjust the fan speed and coolant flow according to the heat dissipation demand predicted by the deep learning model; f2. Establish a communication protocol between the multimodal sensor and the heat dissipation control system; f3. Dynamically adjust the heat dissipation strategy by monitoring the changes in sensor data; f4. Use a data flow control strategy to reduce control system latency and ensure real-time response of the heat dissipation control system.

[0018] Furthermore, in step g, feedback data from the operation of the flying car is collected, and the deep learning model is iteratively trained and optimized using the feedback data, and the following steps are performed: g1. Collecting flying car operation data: Collecting multimodal sensor data from the flying car during operation and actual heat dissipation demand data of the flying car; g2. Using a time series alignment method to match and align the multimodal sensor data and the actual heat dissipation demand data; g3. Preprocessing the aligned data, including outlier processing and standardization; g4. Using iterative training technology to optimize deep learning model parameters, continuously adjusting the internal parameters of the model, and improving the accuracy and robustness of the model; g5. Using the collected operation data in an iterative training manner to continuously optimize the model.

[0019] The advantages of the present invention are: 1. The present invention integrates data collected by multiple sensors and can timely discover potential heat dissipation problems through real-time monitoring, providing an accurate data basis for intelligent heat dissipation management. It combines deep learning models to predict the heat dissipation requirements of flying cars in different working modes, and can automatically generate and implement intelligent heat dissipation strategies that adapt to current working conditions, including adjusting fan speed and coolant flow, to ensure that the flying car can maintain the best heat dissipation state under various working conditions.

[0020] 2. The present invention realizes the automation and intelligence of intelligent heat dissipation management of flying cars through a series of measures such as real-time monitoring, data fusion and standardization, intelligent heat dissipation strategy generation and dynamic adjustment, remote monitoring and optimization, and iterative training and optimization of models. This solution can not only improve the efficiency and reliability of the cooling system of flying cars, but also extend the service life of flying cars, reduce safety hazards caused by poor heat dissipation, and thus enhance the overall flight experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the specific embodiments of the present invention, the drawings required for use in the specific embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 is a flow chart of the method steps of the present invention; DETAILED DESCRIPTION

[0023] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] Embodiment 1: Figure 1 It is a flow chart of the method steps of the present invention, such as Figure 1 As shown, the present invention provides a flying car intelligent heat dissipation management method based on multimodal data, comprising the following steps: a. Real-time monitoring data is collected through a multimodal sensor integrated on the flying car. The multimodal sensor includes a temperature sensor, a pressure sensor, a humidity sensor and an accelerometer. The temperature sensor is installed on the surface of the engine of the flying car to monitor the surface temperature of the engine to ensure that it is within a safe working range. It can be installed inside the battery pack to monitor the internal temperature of the battery pack to prevent safety hazards caused by overheating. It can also be installed at the electronic control unit (ECU) to monitor the temperature of the electronic control unit to ensure its normal operation. The temperature sensor can use a thermocouple, a thermistor or an infrared temperature sensor to collect data at least once a second to ensure real-time monitoring.

[0025] The pressure sensor configuration in the present invention is installed in the hydraulic system of the flying car and on the wing surface and air inlet. It can monitor the pressure distribution on the wing surface of the flying car, which is helpful for understanding the flight status and aerodynamic characteristics. It can also monitor the pressure of the air inlet to ensure smooth engine air intake and avoid performance degradation due to insufficient air intake. The pressure sensor can use a piezoresistive, piezoelectric or strain gauge pressure sensor, and collect data at least once per second.

[0026] The humidity sensors in the present invention are distributed and installed at environmental monitoring points inside and outside the flying car, including the cockpit, engine compartment, battery compartment and external environment, and can monitor humidity changes inside and outside the flying car, thereby evaluating the impact of the environment on the heat dissipation of the flying car. Capacitive, resistive or optical humidity sensors can be used to collect data at least once per second.

[0027] The accelerometer in the present invention is fixedly installed at the center of gravity of the vehicle to monitor the linear and angular acceleration during flight, which helps to monitor the flight status and thermal load changes. The present invention uses a MEMS (micro-electromechanical system) accelerometer, which has the characteristics of high precision and miniaturization, and collects data at least multiple times per second (such as 10 times / second) to capture dynamic changes during flight.

[0028] The collection process of the above sensor data includes: Initialization: After starting the flying car, all sensors enter working state and begin to collect data.

[0029] Data reading: Each sensor reads data according to the set sampling frequency and sends the data to the data acquisition module.

[0030] Timestamp: Add a timestamp to each collected data to ensure the time consistency of sensor data.

[0031] Data transmission: Data is transmitted to the local processing unit via wired or wireless means. For example, CAN bus or Wi-Fi can be used for data transmission.

[0032] Data storage: The local processing unit stores the received data in a local cache for subsequent processing.

[0033] Outlier processing: Use the Z-score method to identify and remove outliers. If the absolute value of the Z-score of a data point is greater than 3, it is considered an outlier and removed. Using the IQR (interquartile range) method, if a data point is less than Q1-1.5IQR or greater than Q3+1.5IQR, it is considered an outlier and removed.

[0034] Missing value processing: For short-term missing values, use forward filling or backward filling methods to fill them. For long-term missing values, use interpolation methods (such as linear interpolation) to fill them.

[0035] Time series alignment: Use timestamps to align all sensor data to ensure that data at the same time point is correctly matched. If the timestamps of some sensor data are inconsistent, they can be aligned through interpolation or synchronization mechanisms.

[0036] Standardization: Use Min-Max scaling to normalize the data to the interval [0,1]. Alternatively, use Z-score standardization to convert the data into a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0037] In summary, the present invention integrates multiple sensors in step a to monitor the key parameters of the flying car in real time, covering monitoring of temperature, pressure, humidity, acceleration and other aspects, so as to timely discover and warn potential safety hazards, such as engine overheating, battery pack temperature being too high, hydraulic system pressure abnormality, etc., to ensure that it is within a safe working range; and provides comprehensive flight status information, providing an accurate and reliable data basis for subsequent data fusion, analysis and intelligent decision-making.

[0038] b. Use data fusion algorithms to fuse data from different sensors, including: b1. Preprocess the sensor data: perform outlier processing on the collected sensor data to remove outliers and missing values ​​in the sensor data; outlier processing can be performed using any two of the following methods: Z-score method: calculate the Z-score of each data point (i.e., the difference between the data point and the mean divided by the standard deviation). If the absolute value of the Z-score of a data point is greater than 3, it is considered an outlier and is removed. IQR method: calculate the interquartile range (IQR), if the data point is less than Q1-1.5IQR or greater than Q3+1.5IQR, it is considered an outlier and is removed. For short-term missing values ​​in the sensor data, use the previous or next valid data point to fill in, and for long-term missing values ​​in the sensor data, use linear interpolation to fill in.

[0039] b2. Use the time series alignment method to match and align different sensor data to ensure the time consistency of sensor data; specifically, the time series alignment method includes: b2-1: timestamp alignment, ensuring that the timestamps of all sensor data are consistent. If the timestamps of some sensor data are inconsistent, align them through interpolation or synchronization mechanisms. b2-2: resampling: resample all sensor data to the same frequency to ensure the time consistency of sensor data.

[0040] b3. Standardization: Ensure that the dimensions and ranges of different sensor data are consistent. Specifically, it includes: b3-1: Use Min-Max scaling: Scale the data to the interval [0, 1]. b3-2: Z-score standardization: Convert the sensor data to a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0041] b4. Correlation analysis and weighted processing of sensor parameters: Calculate the correlation between various sensor parameters and assign weights to comprehensively consider the degree of influence of different parameters on heat dissipation requirements. Specifically, in the intelligent heat dissipation management of flying cars, the degree of influence of different sensor parameters on heat dissipation requirements refers to the specific influence of various sensor data (such as temperature, pressure, humidity and acceleration) on the heat dissipation requirements of flying cars. These parameters affect the thermal load and heat dissipation requirements of flying cars in different ways. Engine surface temperature: directly affects the working state of the engine. High temperature may cause engine performance degradation or even damage. Battery pack internal temperature: Batteries will accelerate aging, reduce efficiency and even pose safety risks at high temperatures. Electronic control unit (ECU) temperature: High temperature will affect the reliability and life of electronic components. Hydraulic system pressure: High pressure may cause the hydraulic system to overheat and increase cooling requirements. Wing surface pressure: The pressure distribution on the wing surface can reflect the flight state. For example, when flying at high speed, the wing surface pressure increases, which may generate more heat. Inlet pressure: Changes in inlet pressure will affect the engine's intake volume, which in turn affects the engine's heat load. Ambient humidity: In a high humidity environment, moisture in the air will increase the burden on the cooling system because water vapor has a large heat capacity and requires more energy to evaporate. Internal humidity: High internal humidity may cause corrosion of electronic equipment and mechanical components, affecting the cooling effect. Linear acceleration: During rapid acceleration or deceleration, the engine and other components will generate more heat. Angular acceleration: When turning or changing attitude, the load distribution of the flying car will change, which may affect the efficiency of the cooling system. The correlation analysis in this step is used to quantify the relationship between various sensor parameters and their relationship with the cooling demand (correlation analysis can be performed using methods including Pearson correlation coefficient and Spearman rank correlation coefficient); the weighted processing in this step is to assign weights to each sensor parameter based on the results of the correlation analysis to comprehensively consider the degree of influence of different parameters on the cooling demand. The weight reflects the importance of each parameter in predicting the cooling demand. Through correlation analysis and weighted processing, the specific degree of influence of each sensor parameter on the cooling demand can be determined, and more effective fusion data can be generated based on this, which helps to improve the prediction accuracy of the subsequent deep learning model and ensure the efficient operation of the flying car under different working conditions.

[0042] c. Build a deep learning model that combines convolutional neural networks with long short-term memory networks to learn the heat dissipation rules of flying cars in different working modes and predict the heat dissipation requirements of flying cars. Specifically, perform the following steps: c1. Collect historical operating data of the flying car, including sensor data and thermal management operation records; c2. Divide the historical operation data into training data set and test data set. Divide the historical operation data into training set (70%), validation set (15%) and test set (15%), ensuring that each data set contains data under various working modes of the flying car (including hovering, cruising, climbing, descending, acceleration, and deceleration) to avoid data imbalance problems.

[0043] c3. Construct a deep learning model based on convolutional neural network and long short-term memory network; In step c3, construct a deep learning model based on convolutional neural network and long short-term memory network, and perform the following steps: c31. Construct an input layer: take multimodal data as the input layer, and define the input layer dimension of the network (define the input layer dimension according to the number of features and time steps of multimodal sensor data); c32. Construct a convolution layer: extract and abstract the input layer data through convolution operation (using multiple convolution kernels to extract local features) and excitation function (using ReLU activation function to increase nonlinearity); c33. Construct a pooling layer: downsample the extracted features (reduce the size of the feature map through maximum pooling or average pooling) to reduce network parameters; c34. Construct a fully connected layer: expand the feature vector output by the pooling layer into a one-dimensional vector and perform a linear transformation; c35. Construct a recurrent layer: use long short-term memory units and gate structures to process time series data, capture contextual information, and realize memory function; c36. Construct an output layer: perform classification or regression operations based on the output of the recurrent layer to generate prediction results.

[0044] c4. Input the training data into the deep learning model for training, and calculate the output prediction value through forward propagation. Specifically, it includes: c41. Divide the original data set into multiple "batches" according to a certain size (such as 32 or 64), randomly select batch data as the input of the model, and randomly select a batch of data to input the model each time for training, instead of processing all the data at once; in deep learning, processing the entire data set at one time (that is, the batch size is equal to the data set size) is called batch gradient descent, but this method consumes a lot of memory and has low computational efficiency. Randomly selecting batch data (i.e., mini-batch gradient descent) can solve these problems to a certain extent. It is not necessary to load all the data into the memory at one time. Only a part of the data needs to be processed each time, which reduces the amount of calculation. Moreover, since a part of the data is randomly extracted for training each time, it can better approach the global optimal solution.

[0045] c42. Perform convolution operation, extract local features through the convolution layer, add the local features to get the hidden state Z. "Hidden State Z" refers to the intermediate representation obtained after adding the local features. It carries the feature information of the input data and can be used as the input of the subsequent layers. In the deep learning model, the hidden state Z is the output of the previous layer and the input of the subsequent layer. Through multi-layer feature extraction and transformation, the original input is converted into a higher-level abstract representation to better complete the task. The hidden state Z is not only the output of the current moment, but also passed to the next time step as part of the memory, enabling the deep learning model to understand and process time series data.

[0046] c43. Use the activation function to transform the hidden state Z nonlinearly to obtain a new hidden state Z′. The nonlinearly transformed hidden state Z′ is used as the input of the next layer and is passed to the deeper part of the network. It plays a connecting role in the deep learning model. It is not only the result of the nonlinear transformation of the current layer, but also the basis for the input of the next layer. Through the nonlinear transformation of the activation function, the deep learning model can capture the complex patterns in the data and construct a more effective feature representation, thereby improving the performance of the deep learning model.

[0047] c44. Perform a pooling operation to reduce the dimension of the hidden state Z′ by downsampling to obtain the output result R of the pooling layer. Specifically, select a window of a fixed size, such as 2x2. Determine the pooling method, take the maximum value in each window as the representative of the area, take the average of all values ​​in each window as the representative of the area, slide the defined pooling window on the feature map, apply the selected pooling method to each window, and obtain a new feature map, that is, the output result R of the pooling layer. Through the pooling operation in step c44, the model can effectively reduce the size of the feature map, improve computational efficiency, and enhance robustness to changes in input data.

[0048] c45. Perform back propagation operation, update weights and biases according to the error between the output result R of the pooling layer and the true value. In step c45, through back propagation, the model adjusts its own parameters according to the prediction error, so that the prediction result is closer to the true value. These two steps are an indispensable part of the deep learning model training process. They work together to enable the model to learn useful information from the training data and gradually improve its prediction ability. Specifically, the value of the loss function is calculated according to the predicted output R and the true label, and the gradient of each weight and bias in the deep learning model is initialized to zero. Starting from the output layer, the gradient of each layer is calculated in turn using the chain rule. Output layer: Calculate the gradient of the output layer. Hidden layer: Calculate the gradient of the current layer according to the gradient of the previous layer. Input layer: Calculate forward until the input layer. Update parameters: Use an optimization algorithm (such as gradient descent, momentum method or Adam) to update the weights and biases of the model according to the calculated gradient. Repeat c41 to c45 until the model converges as expected.

[0049] d. Transmitting the fused data to a remote server or cloud platform for real-time analysis and processing. Specifically, it includes: d1. Transmitting the multimodal sensor data to a remote server or cloud platform via wired or wireless communication; d2. The remote server or cloud platform further analyzes and processes the data, and uses machine learning algorithms or deep learning algorithms to perform multimodal data fusion.

[0050] e. Generate an intelligent cooling strategy for the current working mode of the flying car based on the prediction results of the deep learning model. The intelligent cooling strategy includes adjusting the fan speed and coolant flow. Specifically, it includes the following steps: e1. Set corresponding thresholds for different types of sensor data. By setting the thresholds, the sensor data can be converted into meaningful operation signals, so that the cooling system of the flying car can respond in time when the data exceeds the normal range. e2. Set the fan speed and coolant flow adjustment rules. Specifically, according to the design specifications of the flying car, determine the adjustment range of the fan speed and coolant flow; according to the changes in sensor data, design one or more adjustment strategies. The strategy can be based on a simple proportional controller (P controller) or a more complex PID (proportional-integral-differential) controller. For example, when the temperature rises by 1°C, the fan speed increases by 5%. Multi-mode adaptation: According to different working modes (such as take-off, cruising, landing), set different adjustment rules to ensure that the cooling system of the flying car can adapt to different working conditions; e3. According to the prediction results of the deep learning model, dynamically adjust the fan speed and coolant flow of the flying car.

[0051] f. Implement intelligent cooling strategy and dynamically adjust the cooling strategy according to the actual operation of the flying car. Specifically, it includes: f1. According to the prediction of the cooling demand under the current working mode by the deep learning model, dynamically adjust the fan speed and coolant flow to ensure the effectiveness of the cooling system, including obtaining the prediction results: obtaining the cooling demand prediction under the current working mode from the deep learning model. Calculate the adjustment amount: According to the prediction results, calculate the specific values ​​of the fan speed and coolant flow that need to be adjusted. Send instructions: Convert the calculated adjustment amount into a control instruction and send it to the actuator (such as fan motor, coolant pump, etc.) in the cooling control system. Execute adjustment: The actuator adjusts the fan speed and coolant flow according to the received instructions; f2. Establish a communication protocol between multimodal sensors and cooling control systems to ensure that multimodal sensor data can be efficiently and accurately transmitted to the cooling control system, and realize two-way communication to adjust the cooling strategy in real time, including defining communication standards: Determine the standards of the communication protocol, including data format, transmission frequency, and error checking mechanism. Design interface: Design an interface for sensors and cooling control systems to ensure seamless data exchange between the two. Test the communication link: verify the reliability and stability of the communication link through testing to ensure that data transmission is correct; f3. Dynamically adjust the cooling strategy according to the actual operation status through real-time monitoring of sensor data to adapt to the changing environment and working conditions, including real-time monitoring: continuously monitor the data from multimodal sensors, including temperature, pressure, humidity and other data. Data analysis: analyze the changing trend of sensor data to determine whether the cooling strategy needs to be adjusted. Decision adjustment: dynamically adjust the cooling strategy according to the analysis results, such as changing the fan speed or coolant flow. f4. Reduce the delay in the control system by optimizing the data flow control strategy, ensure that the cooling control system can respond in real time, and improve the response speed and efficiency of the system, including: Priority scheduling: set priority for sensor data to ensure that key data can be processed first. Buffer management: reasonably set the buffer size to avoid data overflow or loss, while reducing the delay of data processing. Queue management: adopt an efficient queue management strategy to ensure that data can be processed in order and avoid data accumulation. Error recovery: design an error recovery mechanism to ensure that data transmission can be quickly restored in the event of a communication failure. Obtain prediction results: obtain the cooling demand prediction under the current working mode from the deep learning model. Calculate the adjustment amount: Based on the prediction results, calculate the specific values ​​of the fan speed and coolant flow that need to be adjusted. Send instructions: Convert the calculated adjustment amount into a control instruction and send it to the actuator in the cooling control system (such as the fan motor, coolant pump, etc.). Execute the adjustment: The actuator adjusts the fan speed and coolant flow according to the received instructions.

[0052] g. Collect feedback data during the operation of the flying car, and use the feedback data to iteratively train and optimize the deep learning model. Specifically, it includes: collecting feedback data during the operation of the flying car, and using the feedback data to iteratively train and optimize the deep learning model. Perform the following steps: g1. Collect flying car operation data: Collect multimodal sensor data and actual heat dissipation demand data of the flying car during operation. Specifically, collect data in real time through multimodal sensors (temperature sensors, pressure sensors, humidity sensors, accelerometers, etc.) on the flying car. Store the collected data, including actual data of sensor readings and heat dissipation requirements. Mark the collected data and record the working mode, environmental conditions and other relevant information at the time. By collecting actual operation data, provide new training samples for the deep learning model, thereby improving the accuracy and robustness of the model. g2. Use the time series alignment method to match and align the multimodal sensor data and the actual heat dissipation demand data. The time series alignment method includes: timestamp recording: recording the timestamp for each sensor data and heat dissipation demand data; time synchronization: ensuring that the timestamps of all sensors are synchronized, and time calibration is performed when necessary; data pairing: pairing the sensor data with the corresponding heat dissipation demand data according to the timestamp; this step can ensure the temporal consistency of the multimodal sensor data and the actual heat dissipation demand data, so as to correctly associate the input features with the output results. g3. Preprocess the aligned data, including outlier processing and standardization. By preprocessing the data, the quality of the data is improved, the impact of noise on model training is reduced, and the data is more suitable for model training; g4. Use iterative training technology to optimize the parameters of the deep learning model, continuously adjust the internal parameters of the model, and improve the accuracy and robustness of the model. Through iterative training, the parameters of the model are continuously optimized to improve the prediction accuracy and robustness of the model. Specifically, it includes splitting the data set: dividing the collected new data set into a training set and a validation set. Parameter adjustment: adjust the model parameters through the back propagation algorithm, use the validation set to evaluate the model performance, and adjust the hyperparameters based on the evaluation results. Model preservation: Save the optimized model for use in subsequent applications. g5. Use the collected operating data to continuously optimize the model through iterative training. Through continuous iterative training, use new operating data to continuously optimize the model to ensure that the model is always up to date. Specifically, it includes: Periodic collection: Regularly collect new operating data, including sensor data and heat dissipation demand data under different working modes. Data integration: Merge new data with existing data sets to form a larger data set. Continuous training: Use the merged data set for model training and continuously iterate and optimize.

[0053] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A flying car intelligent heat dissipation management method based on multimodal data, characterized in that: The following steps are involved: a. Collecting real-time monitoring data through multi-modal sensors integrated on the flying car, the multi-modal sensors include temperature sensors, pressure sensors, humidity sensors and accelerometers; b. Use data fusion algorithm to fuse data from different sensors; c. Build a deep learning model that combines convolutional neural networks with long short-term memory networks to learn the heat dissipation rules of flying cars in different working modes and predict the heat dissipation needs of flying cars; d. Transmit the fused data to a remote server or cloud platform for real-time analysis and processing; e. Generate an intelligent heat dissipation strategy for the current working mode of the flying car based on the prediction results of the deep learning model, wherein the intelligent heat dissipation strategy includes adjusting the fan speed and coolant flow rate; f. Implement intelligent heat dissipation strategies and dynamically adjust the heat dissipation strategies according to the actual operation of the flying car; g. Collect feedback data during the operation of the flying car, and use the feedback data to iteratively train and optimize the deep learning model.

2. The flying car intelligent heat dissipation management method based on multimodal data according to claim 1, characterized in that: The temperature sensors are installed on the engine surface, inside the battery pack and at the electronic control unit of the flying car; the pressure sensors are installed in the hydraulic system, wing surface and air inlet of the flying car; the humidity sensors are installed at the environmental monitoring points inside and outside the flying car; the accelerometer is fixedly installed at the center of gravity of the vehicle to monitor the linear and angular acceleration during flight.

3. The flying car intelligent heat dissipation management method based on multimodal data according to claim 1, characterized in that: In step b, a data fusion algorithm is used to fuse different sensor data, including: b1. Preprocessing the sensor data: performing outlier processing on the collected sensor data to remove outliers and missing values ​​in the sensor data; b2. Using a time series alignment method to match and align different sensor data to ensure the temporal consistency of the sensor data; b3. Standardization processing: ensuring that the dimensions and ranges of different sensor data are consistent; b4. Correlation analysis and weighted processing of sensor parameters: calculating the correlation between each sensor parameter, and assigning weights to comprehensively consider the impact of different parameters on heat dissipation requirements.

4. The flying car intelligent heat dissipation management method based on multimodal data according to claim 1, characterized in that: In step c, a deep learning model is constructed, and the following steps are performed: c1. Collect historical operation data of the flying car, including sensor data and heat dissipation management operation records; c2. Divide the historical operation data into a training data set and a test data set; c3. Construct a deep learning model based on a convolutional neural network and a long short-term memory network; c4. Input the training data into the deep learning model for training, and calculate the output prediction value through forward propagation.

5. The flying car intelligent heat dissipation management method based on multimodal data according to claim 4 is characterized by: In step c3, a deep learning model is constructed based on a convolutional neural network and a long short-term memory network, and the following steps are performed: c31. Construct an input layer: use multimodal data as the input layer, and define the input layer dimension of the network; c32. Construct a convolution layer: extract and abstract features of the input layer data through convolution operations and activation functions; c33. Construct a pooling layer: downsample the extracted features to reduce network parameters; c34. Construct a fully connected layer: expand the feature vector output by the pooling layer into a one-dimensional vector and perform a linear transformation; c35. Construct a recurrent layer: use long short-term memory units and gate structures to process time series data, capture contextual information, and realize memory functions; c36. Construct an output layer: perform classification or regression operations based on the output of the recurrent layer to generate prediction results.

6. The flying car intelligent heat dissipation management method based on multimodal data according to claim 4 is characterized by: In step c4, the training data is input into the deep learning model for training, and the output prediction value is calculated through forward propagation, and the following steps are performed: c41. The original multimodal data set is divided into multiple batches of data, and the batches of data are randomly selected as the input of the model; c42. A convolution operation is performed to extract local features, and the local features are added to obtain the hidden state Z; c43. The hidden state Z is nonlinearly transformed using an activation function to obtain a new hidden state Z′; c44. A pooling operation is performed to reduce the dimension of the hidden state Z′ by downsampling to obtain the pooling layer output result R; c45. A back propagation operation is performed to update the weights and biases according to the error between the pooling layer output result R and the true value, and c41 to c45 are repeated until the convergence of the model reaches the expected value.

7. The flying car intelligent heat dissipation management method based on multimodal data according to claim 1, characterized in that: In step d, the fused data is transmitted to a remote server or cloud platform for real-time analysis and processing, and the following steps are performed: d1. The multimodal sensor data is transmitted to a remote server or cloud platform via wired or wireless communication; d2. The remote server or cloud platform further analyzes and processes the data and uses machine learning algorithms or deep learning algorithms to perform multimodal data fusion.

8. The flying car intelligent heat dissipation management method based on multimodal data according to claim 1, characterized in that: In step e, an intelligent heat dissipation strategy for the current working mode of the flying car is generated according to the prediction results of the deep learning model, and the following steps are performed: e1. Setting corresponding thresholds for different types of sensor data; e2. Setting fan speed and coolant flow adjustment rules; e3. Dynamically adjusting the fan speed and coolant flow of the flying car according to the prediction results of the deep learning model.

9. The flying car intelligent heat dissipation management method based on multimodal data according to claim 1, characterized in that: In step f, an intelligent heat dissipation strategy is implemented, and the heat dissipation strategy is dynamically adjusted according to the actual operation of the flying car, and the following steps are performed: f1. Adjust the fan speed and coolant flow rate according to the heat dissipation demand predicted by the deep learning model; f2. Establish a communication protocol between the multimodal sensor and the heat dissipation control system; f3. Dynamically adjust the cooling strategy by monitoring changes in sensor data; f4. Use data flow control strategy to reduce control system latency and ensure real-time response of the cooling control system.

10. The flying car intelligent heat dissipation management method based on multimodal data according to claim 1, characterized in that: In step g, feedback data from the operation of the flying car is collected, and the deep learning model is iteratively trained and optimized using the feedback data, and the following steps are performed: g1. Collecting flying car operation data: Collecting multimodal sensor data from the flying car during operation and actual heat dissipation demand data of the flying car; g2. Using a time series alignment method to match and align the multimodal sensor data and the actual heat dissipation demand data; g3. Preprocessing the aligned data, including outlier processing and standardization; g4. Using iterative training technology to optimize deep learning model parameters, continuously adjusting the internal parameters of the model, and improving the accuracy and robustness of the model; g5. Using the collected operation data in an iterative training manner to continuously optimize the model.

Images