A dynamic control system and method for a heat-dissipating drum brake based on reinforcement learning

Abstract

The application discloses a kind of based on reinforcement learning's heat dissipation drum brake dynamic control system and method, comprising the following steps: S1, collection and preprocessing drum brake operating data;S2, utilize the neural network of Gated Recurrent Unit and train, output temperature rise trend data;S3, based on proximal policy optimization algorithm training, obtain control strategy network, output heat dissipation control strategy;S4, adopt Harris eagle optimization algorithm, the hyperparameters of heat state prediction model and control strategy network are jointly optimized;S5, according to the model of optimization generation heat dissipation control instruction, control heat dissipation actuator executes corresponding operation;S6, compare temperature feedback data and temperature rise trend data continuously collected, dynamic adjustment is carried out.The application is fused reinforcement learning, deep prediction and intelligent optimization technology, realizes the accurate prediction and dynamic heat dissipation control of drum brake temperature rise trend, to significantly improve the safety, energy efficiency and stability of brake system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of reinforcement learning, and particularly relates to a dynamic control system and method for a heat dissipation drum brake based on reinforcement learning. BACKGROUND

[0002] As an important component of vehicle braking systems, drum brakes are widely used in various passenger cars, commercial vehicles and heavy-duty vehicles, and still dominate in rear-wheel braking and low-speed vehicle systems. Compared with disc brakes, drum brakes have the advantages of compact structure, low cost, good sealing, and easy maintenance. However, under the conditions of long-time braking, frequent braking or high-load operation, the temperature of the drum brake is prone to rise rapidly, and if the heat dissipation is not timely or the control is not reasonable, it is easy to cause heat recession, and further cause braking efficiency decline, thermal expansion deformation, abnormal wear of friction plate and other serious safety problems.

[0003] To deal with the heat recession problem of drum brakes, the existing technology generally uses passive heat dissipation design (such as increasing heat dissipation fins, expanding air vents, improving material thermal conductivity, etc.) or simple active heat dissipation control (such as turning on the fan or liquid cooling device when the temperature exceeds the set threshold) to assist in cooling. However, although passive design can improve the heat dissipation efficiency at the structural level, it is difficult to deal with the sudden situation of rapid heat accumulation during dynamic driving. The active heat dissipation method mostly uses fixed threshold trigger control, which lacks the ability to predict the temperature rise trend, and the response is lagging, which can easily cause the heat dissipation control to be not timely or excessive, affecting the vehicle energy consumption and braking safety.

[0004] In recent years, with the development of sensors, communication and computing platforms, vehicle thermal management systems have begun to evolve towards intelligentization, and some research attempts to introduce model predictive control, fuzzy control, adaptive control and other advanced methods to optimize the control of the heat dissipation process. However, the above methods generally rely on accurate physical modeling and complex control law design, and are not strong in adaptability to changes in driving environment and expansion of data dimensions. For example, model predictive control requires accurate mathematical modeling of vehicle load, speed, braking conditions, etc., which is difficult to maintain long-term stability under actual complex working conditions; and fuzzy control is more subjective in rule design, and is prone to inconsistent control performance when facing high-dimensional dynamic systems.

[0005] In addition, the existing heat dissipation control methods generally have the following technical deficiencies: first, they fail to fully utilize historical operation data to predict temperature rise trends, making it impossible to prepare for heat dissipation in advance and reducing the forward-looking and initiative of the control; second, the control strategies lack intelligent evolution mechanisms and fail to dynamically adjust with changes in working conditions and driving behaviors, resulting in fixed control effects and strong rigidity, which makes it difficult to adapt to complex and changing driving environments; and third, there is a lack of systematic optimization process for the parameters of the heat dissipation control model, and traditional methods often rely on experience to set parameters, making it difficult to obtain optimal performance combinations.

[0006] In recent years, reinforcement learning has made significant progress in the application of automatic control, robot decision-making, energy scheduling, etc. In particular, in the face of systems with large state spaces, multiple control targets, and rapidly changing environments, reinforcement learning has shown strong adaptability and self-learning ability. In the field of vehicle thermal management, reinforcement learning has the natural advantage of learning optimal control strategies through reward and punishment feedback, and can adaptively optimize control strategies through interaction with the environment in the absence of an accurate physical model. However, there are still challenges in directly applying reinforcement learning to drum brake heat dissipation control, such as incomplete state space construction, slow policy convergence speed, and insufficient training samples.

[0007] At the same time, temperature changes are typical time series data, and accurately extracting the trend features of the thermal state over time is the key to achieving accurate prediction. Traditional methods often use simple methods such as moving average and first-order difference for trend analysis, ignoring the long-term dependence information in time series data. The Gated Recurrent Unit neural network (GRU) has become one of the mainstream technologies in the field of time series prediction due to its advantages in processing sequence data, such as simple structure, high computational efficiency, and strong long-term memory capability. Introducing the GRU model into the thermal state prediction of drum brakes can effectively improve the accuracy of temperature rise trend prediction and provide a reliable basis for subsequent control strategies.

[0008] On the other hand, there are a large number of hyperparameters (such as learning rate, number of hidden units, discount factor, reward weight, etc.) in reinforcement learning models and neural network models, which directly affect the model performance and convergence speed. Traditional parameter adjustment methods often rely on human experience or grid search, which is not only inefficient but also prone to local optimization. Therefore, introducing swarm intelligence optimization algorithms (such as particle swarm, differential evolution, whale optimization, etc.) has become an important means to improve parameter optimization efficiency in recent years. Among them, the Harris Hawks Optimization (HHO) algorithm, as a new heuristic intelligent algorithm, simulates the behavior of a hawk group hunting prey, has the characteristics of flexible search strategy, strong local development ability, and excellent global optimization ability, and is very suitable for hyperparameter optimization of high-dimensional nonlinear models.

[0009] However, the application of existing optimization algorithms in heat dissipation systems is mostly limited to parameter optimization of a single model (such as a neural network prediction model), and there is no mechanism for collaborative optimization of prediction and control models, and no consideration of the dynamic influence of external information such as driving mode and working condition characteristics on the optimization process. In addition, existing optimization strategies lack a surprise response mechanism for high-risk thermal state burst situations, and cannot adjust the parameters with large steps in time during the rapid temperature rise stage, limiting the system's ability to quickly adapt to thermal burst events.

[0010] Therefore, how to provide a heat dissipation drum brake dynamic control system and method based on reinforcement learning is a problem that those skilled in the art need to solve. SUMMARY

[0011] One object of the present application is to provide a heat dissipation drum brake dynamic control system and method based on reinforcement learning. The present application fully integrates the time series prediction capability of the gated recurrent unit neural network for thermal state, the intelligent control strategy learning mechanism based on the proximal policy optimization algorithm, and the joint optimization strategy of the improved Harris Hawk optimization algorithm for key model hyperparameters, and describes in detail the whole process of temperature rise trend prediction, heat dissipation strategy dynamic generation and real-time control execution of the drum brake under complex working conditions. The present application has the advantages of high prediction accuracy, fast control response, strong self-adaptation ability and high parameter optimization efficiency, significantly improving the safety and stability of the drum brake under high-frequency use and extreme temperature rise environment.

[0012] According to the heat dissipation drum brake dynamic control method based on reinforcement learning, the method comprises the following steps:

[0013] S1, collecting the running data of the drum brake system, preprocessing the running data to form a thermal state data sequence;

[0014] S2, training the thermal state data sequence using a gated recurrent unit neural network to obtain a thermal state prediction model, and outputting temperature rise trend data;

[0015] S3, based on the thermal state data and the temperature rise trend data, a control strategy network is trained using a proximal policy optimization algorithm, the control strategy network is constructed based on the defined state space and action space, and a reward function is constructed with temperature stability, energy consumption minimization and response time minimization as the target, and a heat dissipation control strategy at each time is outputted;

[0016] S4, the key hyperparameters of the thermal state prediction model and the control strategy network are jointly optimized using a Harris Hawk optimization algorithm to obtain an optimal parameter combination, and an optimized thermal state prediction model and control strategy network are formed;

[0017] S5. Based on the optimized thermal state prediction model and control strategy network, the heat dissipation actuator is controlled to perform corresponding operations according to the output heat dissipation control strategy and the control parameters output by the control strategy network.

[0018] S6. Continuously collect temperature feedback data of the drum brake, compare it with temperature rise trend data, dynamically adjust the parameters of the thermal state prediction model and control strategy network according to the prediction error, and update the training data samples.

[0019] Optionally, the operating data of the drum brake system specifically includes brake temperature, braking duration, brake pedal opening, vehicle speed, ambient temperature, and friction frequency, which are used to generate thermal state data and train the thermal state prediction model and control strategy network.

[0020] Optionally, the preprocessing of the running data specifically includes data cleaning, outlier removal, and normalization, which are used to eliminate noise and unify the data scale, providing standardized input for generating thermal state data.

[0021] Optionally, S2 specifically includes:

[0022] S21. Reassemble the operating data of the drum brake system according to the acquisition time sequence to construct a thermal state data sequence X = {x1, x2, ..., x...} N}, where the i-th thermal state vector is x i , i∈[1,N], where N is the total length of the thermal state data sequence;

[0023] S22. Construct a gated recurrent unit neural network, which includes an input layer, at least one GRU hidden layer, and an output layer, for extracting temporal features from thermal state data sequences and performing temperature prediction modeling.

[0024] S23. In the GRU hidden layer, iteratively update the hidden state at each time step:

[0025] h t =(1-z) t )⊙h t-1 +z t ⊙tanh(Wx t +U(r t ⊙h t-1 )+b);

[0026] Where, x t Let h be the input vector at time t. t Let z be the hidden state vector at time t. t To update the gate vector, r t The gate vector is reset, W is the input weight matrix, U is the recursive weight matrix, b is the bias vector, ⊙ represents element-wise multiplication, and h is the weight vector.t-1 a hidden state vector at time t-1;

[0027] S24, training the gated recurrent unit neural network using the thermal state data sequence X, and optimizing using a mean square error loss function containing a thermal trend consistency term:

[0028]

[0029] wherein y i represents a real temperature value at the i-th moment, represents a predicted temperature value at the i-th moment, λ is a thermal trend consistency penalty coefficient, L represents a total loss function value, N represents the total number of data points in the sample sequence, y i-1 represents a real temperature value at the i-1-th moment, represents a predicted temperature value at the i-1-th moment;

[0030] S25, after the training is completed, obtaining a thermal state prediction model M for modeling the future temperature evolution law of the drum brake;

[0031] S26, inputting the thermal state data sequence X into the thermal state prediction model to obtain temperature rise trend data in a future preset time period.

[0032] Optionally, the S3 specifically comprises:

[0033] S31, combining the thermal state data x t with the temperature rise trend data T′ t and the driving mode data M t to form an extended state vector:

[0034]

[0035] wherein x t represents a thermal state vector at time t, T′ t represents temperature rise trend data at time t, M t represents a driving mode state vector at time t;

[0036] S32, constructing a heat dissipation control action space A = {a1, a2, …, a M}, wherein each action a j represents a heat dissipation control instruction, and M is the total number of actions;

[0037] S33, training a control policy network π by a proximal policy optimization algorithm to establish a mapping relationship between the extended state vector s t and the action a t , π: s t → a t ;

[0038] S34, defining a reward function R t for evaluating the performance of the control strategy:

[0039] R t = -a|T t -T set | -bE t - gT t - d(M t );

[0040] wherein T t represents the actual temperature at time t, T set represents the predetermined temperature, E t represents the energy consumption at time t, T t represents the response time at time t, M t ) represents the driving mode deviation value, a, b, g and d are weight coefficients of temperature deviation, energy consumption, response time and driving mode deviation, respectively;

[0041] S35, in the process of training the control strategy network p, a training target function J(0) is defined by a proximal policy optimization algorithm:

[0042]

[0043] wherein r t (0) represents the policy ratio, represents the advantage function at time t, e represents the clipping parameter, m represents the weight coefficient of the driving mode consistency penalty term, M t ) represents the driving mode deviation at time t, 0 represents the parameter of the control strategy network p, represents the expected value calculation for all time steps t, clip() represents the clipping function, and min represents the minimum value operation;

[0044] S36, after training, the control strategy network p outputs the corresponding heat dissipation control strategy a t for each extended state vector s t .

[0045] Optionally, the S4 specifically comprises:

[0046] S41, determining a key hyperparameter set Q of the thermal state prediction model GRU ;

[0047] S42, determining a key hyperparameter set Q of the control strategy network PPO ;

[0048] S43, constructing a joint hyperparameter set Q = Q GRU U Q PPO, used to uniformly represent all parameters to be optimized in the thermal state prediction model and the control strategy network;

[0049] S44. Construct a physical model for temperature control based on the principles of heat conduction and energy conservation:

[0050] Collect physical parameters related to brake heat conduction, heat dissipation efficiency, and frictional heat generation to form a parameter set θ. p ={k,h,A,m,c}, where k is the thermal conductivity coefficient, h is the convective heat transfer coefficient, A is the heat dissipation area, m is the brake mass, and c is the specific heat capacity;

[0051] A temperature control physical model is constructed based on the principles of energy conservation and heat conduction, and the temperature evolution function T is obtained through discretization. exp (Θ):

[0052] T exp (Θ)=f(T0,θ p );

[0053] Where T0 is the initial temperature, and the function f() is obtained from the heat conduction and convection heat transfer formulas;

[0054] The operating condition index C is constructed based on vehicle operating data, which is composed of braking duration, vehicle speed, and ambient temperature.

[0055] S45. By fusing the training error and physical deviation terms of the temperature control physical model and the control strategy network, the fitness function F(Θ) is defined as the objective evaluation function of the Harris Eagle optimization algorithm:

[0056] F(Θ)=L GRU (Θ GRU )+L PPO (Θ PPO )+α o |T exp (Θ)-T obs |+κ|CC opt |;

[0057] Among them, L GRU (Θ GRU L represents the mean squared prediction error during the training of the control policy network. PPO (Θ PPO T represents the policy error of the control policy network during the training process of the proximal policy optimization algorithm. obs The actual observed temperature, C opt α represents the preset ideal operating condition index. o κ and κ are the weighting factors for the temperature control physical constraint term and the operating condition deviation term, respectively;

[0058] S46, combine the temperature rise trend prediction data and driving mode information, and construct an adaptive dynamic step factor λ through a thermal risk estimation mechanism k :

[0059]

[0060] wherein λ0 represents an initial step size, ΔT is a temperature rise trend, T thr is a temperature warning threshold, T set represents a predetermined temperature, φ(M t ) is a driving mode deviation value, φ ref is a reference driving mode complexity, ζ is a scaling coefficient of driving mode influence, and exp is an exponential function.

[0061] S47, when the temperature rise trend ΔT exceeds the warning threshold T thr , the candidate hyperparameter solution is enhanced and updated through the dynamic step factor λ k :

[0062]

[0063] wherein X (k) represents a candidate hyperparameter combination at the kth iteration, U (k) represents a reference solution obtained based on the current local optimal information, is an indicator function, which takes a value of 1 when ΔT exceeds T thr , and 0 otherwise, and X (k+1) represents a candidate hyperparameter combination at the k+1th iteration.

[0064] S48, under the guidance of the fitness function F(Θ), the candidate hyperparameter set Θ is iteratively searched and updated through the improved Harris Hawk Optimization algorithm, until the preset termination condition is met, so as to obtain the optimal hyperparameter combination Θ * .

[0065] S49, according to the optimal hyperparameter combination Θ * , the parameter of the thermal state prediction model and the control strategy network is updated respectively.

[0066] S410, the updated hyperparameters are used to construct an optimized thermal state prediction model and control strategy network, forming an optimized model under the joint guarantee of temperature control physical constraints, adaptive driving conditions and improved raid mechanism.

[0067] Optionally, the key hyperparameter set of the thermal state prediction model specifically includes the number of units of the GRU hidden layer, the length of the time window, and the learning rate of the GRU network, for adjusting the structural complexity of the thermal state prediction model and the range of time series data processing.

[0068] Optionally, the set of key hyperparameters of the control strategy network specifically comprises a discount factor, a learning rate of the control strategy network, and weight coefficients of temperature bias, energy consumption, response time, and driving mode bias in a reward function, and balances and adjusts the learning rate and reward weight of the control strategy network.

[0069] Optionally, the S5 specifically comprises:

[0070] S51, real-time thermal state data is acquired and input to the optimized thermal state prediction model, and the thermal state prediction model is used to predict the temperature rise trend of the drum brake in a short time in the future, for judging the temperature change trend and control demand;

[0071] S52, the thermal state data at the current time, the predicted temperature rise trend result, and the current driving mode state are input to the optimized control strategy network together;

[0072] S53, the control strategy network outputs a heat dissipation control strategy and control parameters according to the input information, the control strategy comprises heat dissipation mode selection, and the control parameters comprise a start signal of a heat dissipation device, a wind speed level, and a cooling medium flow rate;

[0073] S54, a control instruction is generated according to the control parameters output by the control strategy network, and the control instruction is issued to a corresponding heat dissipation execution mechanism through a communication interface, the heat dissipation execution mechanism comprises a fan, a heat dissipation air duct, or a liquid cooling device;

[0074] S55, after receiving the control instruction, the heat dissipation execution mechanism executes the heat dissipation operation according to the specified parameters, including starting the heat dissipation device, adjusting the wind speed and air volume, and regulating the flow rate of the heat dissipation liquid;

[0075] S56, after completing the control task, the heat dissipation execution mechanism transmits execution state information and real-time temperature feedback data to the system.

[0076] According to the drum brake dynamic control system based on reinforcement learning, the system comprises the following modules:

[0077] A data acquisition and preprocessing module is configured to acquire running data of the drum brake in real time and preprocess the running data;

[0078] A temperature rise prediction module is configured to process the input thermal state data and predict the temperature change trend of the drum brake in a future time period;

[0079] A control strategy generation module is configured to output a heat dissipation control strategy and control parameters according to the thermal state data, the temperature rise trend, and the driving mode state;

[0080] A heat dissipation execution module is configured to receive a control instruction issued by the control strategy generation module and control the running state of a heat dissipation execution mechanism.

[0081] a feedback and update module for receiving the execution state and temperature feedback information returned by the heat dissipation actuator and dynamically updating the parameters of the heat state prediction model and the control strategy network in combination with the prediction error;

[0082] a parameter optimization module for jointly optimizing key hyperparameters in the temperature rise prediction module and the control strategy generation module.

[0083] The present application has the following advantages:

[0084] The dynamic control system and method for heat dissipation drum brake based on reinforcement learning proposed by the present application break through the limitations of the prior art in terms of heat dissipation response lag, rigid control strategy, inaccurate temperature rise prediction, and low parameter tuning efficiency, and achieves multiple beneficial technical effects. By constructing a gated recurrent unit neural network, the present application performs time series modeling and learning on multi-dimensional heat state data during the operation of the drum brake, realizes forward-looking prediction of the temperature rise trend, and overcomes the problem that the traditional moving average or fixed threshold triggering method is difficult to dynamically adapt to heat changes. The control strategy network trained in combination with the proximal policy optimization algorithm enables the system to generate the optimal heat dissipation control strategy and specific parameters in real time according to different heat states, driving conditions, and temperature rise trends, thereby realizing more intelligent and efficient heat dissipation control decisions.

[0085] In terms of optimization, the present application introduces a temperature control physical model constructed based on heat conduction and convective heat transfer theory, and uses it as part of the fitness evaluation, in combination with the improved Harris Hawk optimization algorithm, to realize joint optimization of the key hyperparameters of the heat state prediction model and the control strategy network. In particular, when the temperature rise trend prediction result exceeds the warning threshold, the dynamic step adjustment mechanism guides the candidate solution to perform enhanced raid updates, thereby significantly improving the response speed and global search ability of the algorithm in high temperature burst situations, ensuring that the system can output the optimal heat dissipation strategy in time to avoid the risk of heat recession.

[0086] In addition, the system structure designed by the present application covers complete closed-loop control processes such as data acquisition, preprocessing, prediction, control, optimization, and feedback, and has high integration and engineering adaptability. The system can continuously and stably operate under various complex conditions, and has good self-learning and self-adaptive abilities. Overall, the present application realizes precise modeling and active control of the temperature rise process of the drum brake, improves the intelligent level of the thermal management of the vehicle braking system, and has significant engineering application value in improving braking safety, prolonging the service life of parts, and reducing energy loss. BRIEF DESCRIPTION OF DRAWINGS

[0087] The accompanying drawings are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and are meant to explain the present application but are not intended to limit the application. In the drawings:

[0088] Figure 1 A flow chart of a dynamic control method of a heat dissipation drum brake based on reinforcement learning is proposed for the present application;

[0089] Figure 2 A structural schematic diagram of a dynamic control system of a heat dissipation drum brake based on reinforcement learning is proposed for the present application. DETAILED DESCRIPTION

[0090] The present application will now be further described in detail with reference to the drawings. These drawings are simplified schematic diagrams and only schematically illustrate the basic structure of the present application, and thus only show the components related to the present application.

[0091] REFERENCE Figure 1 A dynamic control method of a heat dissipation drum brake based on reinforcement learning, comprising the following steps:

[0092] S1, collecting the operation data of the drum brake system, pre-processing the operation data to form a heat state data sequence;

[0093] S2, training the heat state data sequence using a gated recurrent unit neural network to obtain a heat state prediction model, and outputting temperature rise trend data;

[0094] S3, based on the heat state data and the temperature rise trend data, training a control strategy network using a proximal policy optimization algorithm, the control strategy network being based on defined state space and action space, and constructing a reward function with temperature stability, minimum energy consumption and shortest response time as the target, and outputting a heat dissipation control strategy at each time;

[0095] S4, jointly optimizing the key hyperparameters of the heat state prediction model and the control strategy network using a Harris hawk optimization algorithm to obtain an optimal parameter combination, and forming an optimized heat state prediction model and control strategy network;

[0096] S5, based on the optimized heat state prediction model and control strategy network, controlling the heat dissipation execution mechanism to perform corresponding operations according to the output heat dissipation control strategy and the control parameters output by the control strategy network;

[0097] S6, continuously collecting temperature feedback data of the drum brake, comparing the temperature feedback data with the temperature rise trend data, dynamically adjusting the parameters of the heat state prediction model and the control strategy network according to the prediction error, and updating the training data samples.

[0098] The application realizes accurate prediction of temperature rising trend and intelligent generation of real-time heat dissipation strategy by using reinforcement learning technology to dynamically control the heat dissipation process of the drum brake. High-quality thermal state data sequences are obtained by using a data acquisition and preprocessing module, and the sequences are trained by a gated recurrent unit neural network. The system can accurately predict future temperature trends, avoiding the response lag and temperature surge risk of traditional fixed threshold control methods. The control strategy network trained by the proximal policy optimization algorithm generates optimal heat dissipation control strategies in a multi-dimensional state space, ensuring that the heat dissipation device starts at the best time, thereby reducing energy consumption and improving brake safety. The improved Harris Hawk optimization algorithm is used to jointly optimize the key hyperparameters of the temperature rising prediction model and the control strategy network, so that the entire system can quickly adjust the control strategy and achieve adaptive closed-loop control when facing complex working conditions and sudden high temperatures. The system continuously collects temperature feedback data and compares it with the predicted temperature rising trend, dynamically adjusts the model parameters according to the prediction error, and continuously improves the heat dissipation control accuracy. In summary, the application has the advantages of accurate temperature rising prediction, fast response, low energy consumption, flexible control strategy and strong adaptive ability, significantly improving the heat dissipation performance and safety of the drum brake, and is suitable for practical applications under various complex working conditions.

[0099] In the embodiment, the operating data of the drum brake system specifically includes brake temperature, brake duration, brake pedal opening, vehicle speed, ambient temperature and friction frequency, which are used to generate thermal state data, train a thermal state prediction model and a control strategy network.

[0100] In the embodiment, the operating data preprocessing specifically includes data cleaning, outlier removal and normalization processing, which are used to eliminate noise and unify data scales, providing standardized input for generating thermal state data.

[0101] In the embodiment, S2 specifically includes:

[0102] S21, reorganize the operating data of the drum brake system according to the collection time sequence, and construct a thermal state data sequence X={x1, x2,…, x N} where the i-th thermal state vector is x i , i∈[1,N], N is the total length of the thermal state data sequence;

[0103] S22, construct a gated recurrent unit neural network, which includes an input layer, at least one GRU hidden layer and an output layer, for extracting time sequence features in the thermal state data sequence and modeling temperature prediction;

[0104] S23, in the GRU hidden layer, the hidden state of each time is iteratively updated:

[0105] h t =(1-z t )⊙h t-1 +z t ⊙tanh(Wx t +U(r t ⊙h t-1 )+b);

[0106] where x t is the input vector at time t, h t is the hidden state vector at time t, z t is the update gate vector, r t is the reset gate vector, W is the input weight matrix, U is the recurrent weight matrix, b is the bias vector, represents element-wise multiplication, h t-1 represents the hidden state vector at time t-1.

[0107] S24, train the gated recurrent unit neural network using the thermal state data sequence X, and optimize using a mean square error loss function containing a thermal trend consistency term:

[0108]

[0109] where y i represents the true temperature value at time i, y represents the predicted temperature value at time i, λ is the thermal trend consistency penalty coefficient, L represents the total loss function value, N represents the total number of data points in the sample sequence, y i-1 represents the true temperature value at time i-1, y represents the predicted temperature value at time i-1.

[0110] S25, after training, obtain the thermal state prediction model M, which is used to model the future temperature evolution law of the drum brake;

[0111] S26, input the thermal state data sequence X to the thermal state prediction model to obtain the temperature rise trend data in the future preset time period.

[0112] The application utilizes a gated recurrent unit neural network to perform time series modeling on the operation data of a drum brake, reorganizes the data in chronological order to form a thermal state data sequence, and realizes comprehensive utilization of multi-dimensional data such as brake temperature, braking duration, pedal opening, vehicle speed, ambient temperature and friction frequency. On this basis, a mean square error loss function containing a thermal trend consistency term is used to train the neural network, effectively reducing the prediction error caused by data noise and environmental changes, and ensuring the accuracy and robustness of temperature prediction. The thermal state prediction model obtained after training can accurately capture the dynamic law of future temperature evolution of the drum brake, and output the temperature rise trend data in advance, providing reliable forward-looking input for the subsequent heat dissipation control strategy based on reinforcement learning. This method not only realizes the goal of real-time prediction of temperature change, but also has the advantages of fast response speed, high prediction accuracy and strong adaptability, thereby effectively preventing the risk of thermal recession under high-frequency braking and complex working conditions, and improving the overall safety and durability of the system.

[0113] In the embodiment, the S3 specifically includes:

[0114] S31, combining the thermal state data x t , the temperature rise trend data T' t and the driving mode data M t to form an extended state vector:

[0115]

[0116] wherein x t represents the thermal state vector at time t, T' t represents the temperature rise trend data at time t, M t represents the driving mode state vector at time t;

[0117] S32, constructing a heat dissipation control action space A={a1, a2, …, a M}, wherein each action a j represents a heat dissipation control instruction, and M is the total number of actions;

[0118] S33, training the control policy network π by a proximal policy optimization algorithm to establish a mapping relationship between the extended state vector s t and the action a t , π:s t →a t ;

[0119] S34, defining a reward function R t for evaluating the performance of the control policy:

[0120] R t = -α|T t -Tset | -βE t -γτ t -δφ(M t );

[0121] wherein, T t represents the actual temperature at time t, T set represents the predetermined temperature, E t represents the energy consumption at time t, τ t represents the response time at time t, φ(M t ) represents the driving mode deviation value, and α, β, γ and δ are weight coefficients of the temperature deviation, the energy consumption, the response time and the driving mode deviation, respectively;

[0122] S35. In the process of training the control policy network π, a training target function J(θ) is defined by a proximal policy optimization algorithm:

[0123]

[0124] wherein, r t (θ) represents the policy ratio, represents the advantage function at time t, ε represents a clipping parameter, μ represents a weight coefficient of the driving mode consistency penalty term, Δ(M t ) represents the driving mode deviation amount at time t, and θ represents the parameters of the control policy network π, represents an expected value calculation for all time steps t, clip() represents a clipping function, and min represents a minimum value operation;

[0125] S36. After the training is completed, the control policy network π outputs a corresponding heat dissipation control policy a t for each extended state vector s t .

[0126] The present application realizes multi-dimensional perception and expression of the current operating state of the drum brake by constructing an extended state vector containing thermal state, temperature rise trend and driving mode information, and enhances the context relevance of the control strategy generation. On this basis, by constructing a heat dissipation control action space and training the control strategy network using the proximal policy optimization algorithm, an efficient mapping relationship from the extended state vector to the optimal heat dissipation control action is established, significantly improving the stability and convergence speed of the strategy generation. By designing a multi-objective reward function, temperature stability, energy consumption level, response time and driving mode deviation are taken as comprehensive evaluation criteria to ensure that the heat dissipation control strategy of the system under complex working conditions is not only accurate and effective, but also takes into account energy efficiency and comfort. Further, in the strategy training process, a clipping function and driving mode consistency constraint term are introduced to effectively avoid unstable behavior caused by excessive strategy update and improve the matching degree of the control strategy and actual driving behavior. After training, the control strategy network can output real-time, fine-grained and dynamic heat dissipation control strategies according to the extended state vector at any time, realizing feedforward control and real-time adjustment of temperature rise. Overall, this method has the advantages of strong strategy stability, high control accuracy and strong self-adaptability, and can be widely applied in intelligent vehicle thermal management systems.

[0127] In the present embodiment, S4 specifically includes:

[0128] S41, determining a key hyperparameter set Θ of the thermal state prediction model GRU ;

[0129] S42, determining a key hyperparameter set Θ of the control strategy network PPO ;

[0130] S43, constructing a joint hyperparameter set Θ = Θ GRU ∪ Θ PPO , used to uniformly represent all parameters to be optimized of the thermal state prediction model and the control strategy network;

[0131] S44, constructing a temperature control physical model based on heat conduction and energy conservation principles:

[0132] Collecting physical parameters related to brake heat conduction, heat dissipation efficiency and friction heat generation to form a parameter set θ p ={k, h, A, m, c}, where k is the thermal conductivity coefficient, h is the convective heat transfer coefficient, A is the heat dissipation area, m is the brake mass, and c is the specific heat capacity;

[0133] Based on the energy conservation and heat conduction principles, a temperature control physical model is constructed, and the temperature evolution function T exp (Θ) is obtained by discretization:

[0134] T exp (Θ) = f(T0, θp );

[0135] Wherein, T0 is the initial temperature, and the function f() is obtained from the heat conduction and convection heat transfer formula;

[0136] A working condition index C is constructed according to the vehicle operation data, wherein the working condition index is composed of the brake duration, the vehicle speed and the ambient temperature;

[0137] S45, define the fitness function F(Θ) as the target evaluation function of the Harris Hawk optimization algorithm by fusing the training error of the temperature control physical model and the control strategy network and the physical deviation term:

[0138] F(Θ)=L GRU (Θ GRU )+L PPO (Θ PPO )+α o |T exp (Θ)-T obs |+κ|C-C opt |;

[0139] Wherein, L GRU (Θ GRU ) represents the mean square prediction error in the training process of the control strategy network, L PPO (Θ PPO ) represents the strategy error of the control strategy network in the training process of the proximal policy optimization algorithm, T obs represents the observed actual temperature, C opt represents the preset ideal working condition index, and α o and κ are weight factors of the temperature control physical constraint term and the working condition deviation term, respectively;

[0140] S46, combine the temperature rise trend prediction data and the driving mode information, and construct an adaptive dynamic step factor λ k through a thermal risk estimation mechanism:

[0141]

[0142] Wherein, λ0 represents the initial step length, ΔT is the temperature rise trend, T thr is the temperature warning threshold, T set represents the predetermined temperature, φ(M t ) is a driving mode deviation value, φ ref is the reference driving mode complexity, ζ is the scaling coefficient of the driving mode influence, and exp is the exponential function;

[0143] S47, when the temperature rise trend ΔT exceeds the warning threshold T thr , the candidate hyperparameter solution is enhanced and updated by the dynamic step factor λ k :

[0144]

[0145] wherein X (k) represents the candidate hyperparameter combination at the kth iteration, U (k) represents the reference solution obtained based on the current local optimal information, is an indicator function, which takes the value 1 when ΔT exceeds T thr , and 0 otherwise, X (k+1) represents the candidate hyperparameter combination at the k+1th iteration;

[0146] S48, under the guidance of the fitness function F(Θ), the candidate hyperparameter set Θ is iteratively searched and updated by the improved Harris Hawk Optimization algorithm until the preset termination condition is met, so as to obtain the optimal hyperparameter combination Θ * ;

[0147] S49, according to the optimal hyperparameter combination Θ * , the parameter updating is performed on the thermal state prediction model and the control strategy network respectively;

[0148] S410, the updated hyperparameters are used to construct the optimized thermal state prediction model and control strategy network, forming an optimized model under the joint guarantee of temperature control physical constraints, adaptive driving conditions and improved raid mechanism.

[0149] The improved Harris Hawk Optimization algorithm is introduced to realize the collaborative optimization of the key hyperparameters of the thermal state prediction model and the control strategy network, and the problems of low efficiency, poor global search ability and difficulty in adapting to dynamic conditions in the traditional manual parameter setting or single model optimization method are solved. First, the temperature control physical model is constructed by the heat conduction and energy conservation principle, the key physical parameters such as heat conduction coefficient, heat exchange area and mass are introduced, and the indicators reflecting the actual working conditions are constructed combined with the vehicle operation data, so that the optimization process can evaluate the model performance under the premise of meeting the physical constraints and working condition adaptability. Secondly, a comprehensive fitness function including prediction error, strategy error, physical deviation and working condition deviation is constructed as the optimization target, so that the algorithm can optimize the prediction accuracy, strategy reliability and system adaptability at the same time. In the optimization strategy, the adaptive dynamic step factor is constructed combined with the temperature rise trend prediction result and the driving mode deviation, and the enhanced raid update is triggered when the high temperature risk appears, which accelerates the algorithm convergence and improves the response ability to sudden risks. Finally, the joint parameter updating is performed on the thermal state prediction model and the control strategy network, and the intelligent thermal control system integrating physical mechanism and data-driven is constructed. This method effectively improves the scientificity and efficiency of parameter tuning, and has the significant advantages of stable control effect, reliable optimization result and strong adaptability to working conditions.

[0150] In this embodiment, the key hyperparameter set of the thermal state prediction model specifically includes the number of units of the GRU hidden layer, the length of the time window, and the learning rate of the GRU network, which is used to adjust the structural complexity of the thermal state prediction model and the processing range of the time series data.

[0151] In this embodiment, the key hyperparameter set of the control strategy network specifically includes the discount factor, the learning rate of the control strategy network, and the weight coefficients of the temperature deviation, energy consumption, response time, and driving mode deviation in the reward function, which balances and adjusts the learning rate and reward weight of the control strategy network.

[0152] In this embodiment, the S5 specifically includes:

[0153] S51, acquiring real-time thermal state data and inputting it to the optimized thermal state prediction model, using the thermal state prediction model to predict the temperature rise trend of the drum brake in the future short time, for judging the temperature change trend and control demand;

[0154] S52, inputting the thermal state data at the current time, the predicted temperature rise trend result, and the current driving mode state to the optimized control strategy network;

[0155] S53, the control strategy network outputs the heat dissipation control strategy and control parameters according to the input information, the control strategy includes heat dissipation mode selection, and the control parameters include the start signal of the heat dissipation device, the wind speed level, and the cooling medium flow rate;

[0156] S54, generating a control instruction according to the control parameters output by the control strategy network, and issuing the control instruction to the corresponding heat dissipation execution mechanism through a communication interface, the heat dissipation execution mechanism includes a fan, a heat dissipation air duct, or a liquid cooling device;

[0157] S55, after receiving the control instruction, the heat dissipation execution mechanism executes the heat dissipation operation according to the specified parameters, including starting the heat dissipation device, adjusting the wind speed and air volume, and regulating the flow rate of the heat dissipation liquid;

[0158] S56, after completing the control task, the heat dissipation execution mechanism transmits the execution state information and real-time temperature feedback data to the system.

[0159] The application realizes intelligent execution and dynamic adjustment of the heat dissipation process of the drum brake by constructing a closed-loop real-time control mechanism, and overcomes problems such as response lag, control rigidity and feedback loss of the traditional system. After obtaining the thermal state data, the system immediately calls the optimized thermal state prediction model to predict the temperature rise trend in the future short time, thereby sensing the temperature change risk in advance and providing feedforward support for subsequent control. By inputting the current thermal state, temperature rise trend and driving mode into the optimized control strategy network, the system can output a heat dissipation control strategy and parameters highly matched with the current working condition, and realize fine regulation and control of multiple control variables such as fan start, wind speed level, cooling medium flow rate and the like. The control command is quickly issued to the fan, air duct or liquid cooling and other heat dissipation execution mechanisms through the communication module, ensuring that the command response is timely and the execution action is accurate. After the heat dissipation execution is completed, the system can also obtain real-time temperature feedback and execution state information, providing closed-loop verification and dynamic correction basis for the prediction model and strategy network. The method has the advantages of fast response speed, accurate control strategy, complete feedback mechanism and strong self-adaptive ability, and can significantly improve the temperature control safety and system operation stability of the drum brake under high-frequency braking and complex working conditions.

[0160] According to the embodiment of the application, a dynamic control system of a heat dissipation drum brake based on reinforcement learning comprises the following modules:

[0161] A data acquisition and preprocessing module is configured to acquire running data of the drum brake in real time and preprocess the running data.

[0162] A temperature rise prediction module is configured to process the input thermal state data and predict the temperature change trend of the drum brake in a future time period.

[0163] A control strategy generation module is configured to output a heat dissipation control strategy and control parameters according to the thermal state data, temperature rise trend and driving mode state.

[0164] A heat dissipation execution module is configured to receive the control instructions issued by the control strategy generation module and control the running state of the heat dissipation execution mechanism.

[0165] A feedback and updating module is configured to receive the execution state and temperature feedback information returned by the heat dissipation execution mechanism, and dynamically update the parameters of the thermal state prediction model and the control strategy network in combination with the prediction error.

[0166] A parameter optimization module is configured to jointly optimize the key hyperparameters in the temperature rise prediction module and the control strategy generation module.

[0167] Embodiment 1:

[0168] In order to verify the feasibility of the application in implementation, the application is applied to a certain brand of light truck, which runs on a long downhill section in mountainous area. The total length of the section is about 21 kilometers, including many long slopes, sharp curves, tunnels and frequent braking conditions, which are extremely risky for heat recession. The test vehicle is a rear-wheel drive light truck equipped with drum brakes, with a full load of 4.5 tons. The braking system is drum rear wheel + liquid-cooled auxiliary heat structure. The vehicle is equipped with a standard temperature acquisition system, an ECU control unit, an intelligent fan execution module and a braking system data analysis platform.

[0169] Under the traditional system, this type of vehicle usually uses a fixed threshold to trigger the fan for heat dissipation control, that is, when the brake drum temperature exceeds a certain set value (such as 220℃), the heat dissipation device is automatically turned on. However, this method has obvious defects: control response is lagging, strategy is single, energy consumption is large, and it is easy to appear repeated high temperature or even heat recession risk under complex slope.

[0170] In this embodiment, the system of the application is integrated into the test vehicle, which is connected to the vehicle CAN bus through the front deployment form, and the intelligent prediction of drum brake temperature rise trend, dynamic control strategy generation and parameter optimization modules are enabled. During actual operation, the system continuously collects the running state information of the vehicle, including vehicle speed, braking intensity, duration, environmental temperature and friction frequency, etc. multidimensional signals, forming a complete heat state vector and inputting it into the optimized gate cycle unit neural network. The network output is the temperature rise trend data in the next 60 seconds, which is used to generate the current optimal heat dissipation control instruction by the control strategy network.

[0171] The control strategy network is trained based on the proximal policy optimization algorithm, which can dynamically generate control parameters such as fan start level and liquid cooling medium flow rate adjustment value under the multi-objective function considering temperature stability, response speed, energy efficiency and driving mode matching degree. The system also embeds an improved Harris Hawk optimization module, which immediately calls the adaptive surprise strategy to fine-tune the key hyperparameters online when the system detects a sharp fluctuation in temperature rise trend or high temperature risk trigger, ensuring the optimal response capability of the control strategy under sudden state.

[0172] In the actual road test for 6 consecutive days, the control system of the application (experimental group) and the traditional threshold control system (comparison group) are respectively run on the same route, and the key indicators such as temperature peak value, fan response time, energy consumption of heat dissipation process, etc. are tested during each braking process.

[0173] Table 1 Comparison of drum brake heat dissipation performance data

[0174]

[0175]

[0176] From the data in Table 1, it can be seen that the reinforcement learning driven heat dissipation control system proposed in the application is significantly better than the traditional fixed threshold control method in many key performance indicators. In terms of average maximum temperature, the brake drum temperature of the experimental group is effectively controlled at 187.3℃, while the comparison group reaches 226.8℃, with a reduction of 17.4%. This result shows that the system can effectively suppress temperature peaks and avoid performance degradation or thermal recession problems due to overheating of the braking system. Further analysis of the number of times the temperature exceeds 220℃ shows that the experimental group did not have a temperature overrun event in continuous multi-day, multi-downhill tests, while the traditional control system had 12 times when the temperature exceeded 220℃. The results show that the system has strong high-temperature risk prevention ability by predicting temperature trends in advance and dynamically adjusting the heat dissipation strategy, significantly improving the safety and reliability of the system. In terms of control response, the average response delay of the fan is also an important indicator of the intelligence of the system. The response time of the experimental group is 0.46 seconds, which is much faster than the 2.13 seconds of the comparison group, with an advance response of 78.4%. This shows that the system intervenes in control when the temperature trend is initially emerging through the real-time strategy generation mechanism, avoiding the lag of traditional "after-the-fact" control.

[0177] From the energy consumption point of view, the average fan opening time of the experimental group is 148 seconds, which is more than 50 seconds less than the 201 seconds of the comparison group, saving about 26.4% of the heat dissipation energy. This shows that on the premise of achieving better temperature control, the application can also reduce the fan workload, improve overall energy efficiency, and prolong the service life of the equipment.

[0178] Finally, in terms of heat dissipation efficiency, the time it takes for the temperature to fall from the high point to the set safety range is an indicator that reflects the responsiveness of the model and the control ability of the system. The experimental group only needs an average of 176 seconds to complete the falling process, while the comparison group needs 292 seconds, with an improvement of 40%. This shows that in the closed-loop execution of heat dissipation control, the system not only responds faster and controls more accurately, but also has more efficiency in recovering cooling.

[0179] In summary, the table data fully reflects the technical advantages of the system in the field of drum brake heat dissipation control. By integrating reinforcement learning, gated recurrent unit network and improved Harris hawk optimization algorithm, the system realizes intelligent closed-loop management from prediction to control, successfully solves the problems of response lag, high-temperature loss of control and high energy consumption in traditional heat dissipation control methods, and provides a feasible and efficient engineering path for heat management intelligence.

[0180] The above merely describes preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art, according to the technical solution and inventive concept of the present application, makes equivalent replacement or change within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A dynamic control method of a heat dissipating drum brake based on reinforcement learning, characterized by, The method comprises the following steps: S1, collecting the operation data of the drum brake system, preprocessing the operation data to form a thermal state data sequence; S2, training the thermal state data sequence using a gated recurrent unit neural network to obtain a thermal state prediction model, and outputting temperature rise trend data; S3, based on the thermal state data and the temperature rise trend data, a control strategy network is trained using a proximal policy optimization algorithm, the control strategy network is constructed based on the defined state space and action space, a reward function is constructed with temperature stability, energy consumption minimization and response time minimization as the target, and a heat dissipation control strategy at each time is outputted; S4, the key hyperparameters of the thermal state prediction model and the control strategy network are jointly optimized using a Harris hawk optimization algorithm to obtain an optimal parameter combination, and an optimized thermal state prediction model and control strategy network are formed; S5, heat dissipation control instructions are generated based on the optimized thermal state prediction model and the control strategy network, and the heat dissipation execution mechanism is controlled to perform corresponding operations; S6, the temperature feedback data of the drum brake is continuously collected, compared with the temperature rise trend data, the parameters of the thermal state prediction model and the control strategy network are dynamically adjusted according to the prediction error, and the training data samples are updated; The S2 specifically comprises: S21, reorganize the operation data of the drum brake system according to the collection time sequence, construct a thermal state data sequence X = {X 1, X2…,X N}, wherein the i th thermal state vector is X i , i\in \left [ {1,N} \right ] , and N is the total length of the thermal state data sequence; S22, a gated recurrent unit neural network is constructed, the gated recurrent unit neural network comprises an input layer, at least one GRU hidden layer and an output layer, and is used for extracting time sequence features in the thermal state data sequence and performing temperature prediction modeling; S23, in the GRU hidden layer, the hidden state at each time is iteratively updated; S24, training the gated recurrent unit neural network using the thermal state data sequence, optimizing using a mean squared error loss function including a thermal trend consistency term: ; wherein y i represents the real temperature value at the i-th moment, represents the predicted temperature value at the i-th moment, is a thermal trend consistency penalty coefficient, L represents a total loss function value, N represents the total number of data points in a sample sequence, y i-1 represents the real temperature value at the i-1-th moment, represents the predicted temperature value at the i-1-th moment; S25, after the training is completed, a thermal state prediction model M is obtained, which is used for modeling the future temperature evolution law of the drum brake; S26, the thermal state data sequence is inputted into the thermal state prediction model to obtain temperature rise trend data in a future preset time period.

2. The dynamic control method of a heat-dissipation drum brake based on reinforcement learning according to claim 1, characterized in that, The operation data of the drum brake system specifically comprises brake temperature, brake duration, brake pedal opening, vehicle speed, ambient temperature and friction frequency, which is used for generating thermal state data, training the thermal state prediction model and the control strategy network.

3. The dynamic control method of a heat-dissipation drum brake based on reinforcement learning according to claim 1, characterized in that, The operation data preprocessing specifically comprises data cleaning, outlier rejection and normalization processing, which is used for eliminating noise and unifying data scales to provide standardized input for generating thermal state data.

4. The dynamic control method of a heat-dissipation drum brake based on reinforcement learning according to claim 1, characterized in that, The S3 specifically comprises: S31, combine the thermal state data x t with the temperature rise trend data and the driving mode data M t to form an extended state vector: ; wherein x t represents a thermal state vector at time t, represents temperature rise tendency data at time t, M t represents a driving mode state vector at time t; S32, construct heat dissipation control action space A = {a 1, a 2, …,a m}, wherein each action a j represents a heat dissipation control instruction, and M is the total number of actions; S33, training the control policy network through a proximal policy optimization algorithm , establishing an extended state vector s t and a mapping relationship between the state s t and the action a ; S34, define a reward function for evaluating the performance of the control policy: ; wherein T t represents the actual temperature at time t, T set represents a predetermined temperature, E t represents the energy consumption at time t, represents the response time at time t, represents the driving mode deviation value, , , and are weight coefficients of the temperature deviation, the energy consumption, the response time and the driving mode deviation, respectively. S35、in the process of training the control strategy network, define a training objective function through a proximal policy optimization algorithm : ； wherein, denotes a policy ratio, denotes a advantage function at time t, denotes a clipping parameter, denotes a weight coefficient of the driving mode consistency penalty term, denotes a driving mode deviation at time t, denotes a parameter of the control policy network , denotes an expected value calculation over all time steps t, clip() denotes a clipping function, and min denotes a minimum operation. S36、After the training is completed, the control policy network for each extended state vector output the corresponding heat dissipation control policy .

5. The dynamic control method of a heat-dissipation drum brake based on reinforcement learning according to claim 1, characterized in that, The S4 specifically comprises: S41, determine the key hyperparameter set of the thermal state prediction model ; S42, determining a set of key hyperparameters of the control policy network ; S43, constructing a joint hyperparameter set for uniformly representing all the parameters to be optimized of the thermal state prediction model and the control strategy network; S44, a temperature control physical model is constructed by heat conduction and energy conservation principles: Collecting physical parameters related to brake heat conduction, heat dissipation efficiency and friction heat generation to form a parameter set wherein k is the thermal conductivity coefficient, h is the convective heat transfer coefficient, A is the heat dissipation area, m is the brake mass, and c is the specific heat capacity. A temperature control physical model is constructed based on the energy conservation and heat conduction principle, and a temperature evolution function is obtained by discretization solution : ； Wherein, T0 is the initial temperature, and the function f() is obtained from the heat conduction and convection heat transfer formula; A working condition index C is constructed according to the vehicle operation data, wherein the working condition index is composed of brake duration, vehicle speed and ambient temperature; S45, define fitness function by fusing training error of temperature control physical model and control strategy network with physical deviation term As the objective evaluation function of Harris Hawk optimization algorithm: ； wherein, represents the mean square prediction error in the control policy network training process, represents the policy error in the proximal policy optimization algorithm training process of the control policy network, represents the observed actual temperature, represents the preset ideal working condition index, and are the weight factors of the temperature control physical constraint term and the working condition deviation term, respectively. S46、 Combining the temperature rise trend prediction data and the driving mode information, an adaptive dynamic step factor is constructed through a thermal risk estimation mechanism : ； wherein, denotes an initial step size, is a temperature trend, is a temperature warning threshold, denotes a predetermined temperature, denotes a driving mode deviation value, is a reference driving mode complexity, is a scaling factor for driving mode influence, is an exponential function; S47, when the temperature rise trend Exceeding the warning threshold by a dynamic step factor Enhanced raid update on candidate hyperparameter solution: ； wherein, denotes the candidate hyperparameter combination at the kth iteration, denotes the reference solution based on the current local optimum information, is an indicator function, when exceeds is 1, otherwise 0, denotes the candidate hyperparameter combination at the k+1th iteration; S48. By using the improved Harris Eagle optimization algorithm, in the fitness function Under the guidance of the candidate hyperparameter set The process involves iterative searching and updating until a preset termination condition is met, thereby obtaining the optimal combination of hyperparameters. ; S49、according to the optimal hyperparameter combination The parameter updating is respectively performed on the thermal state prediction model and the control strategy network. S410, the optimized thermal state prediction model and the control strategy network are constructed using the updated hyperparameters to form an optimization model under the joint guarantee of temperature control physical constraints, adaptive driving working conditions and improved attack mechanism.

6. The dynamic control method of a heat-dissipation drum brake based on reinforcement learning according to claim 5, characterized in that, The set of key hyperparameters of the thermal state prediction model specifically comprises the number of units of the GRU hidden layer, the length of the time window and the learning rate of the GRU network, which is used for adjusting the structure complexity of the thermal state prediction model and the time sequence data processing range.

7. The dynamic control method of a heat-dissipation drum brake based on reinforcement learning according to claim 6, characterized in that, The key hyperparameter set of the control strategy network specifically includes a discount factor, a learning rate of the control strategy network, and weight coefficients of temperature bias, energy consumption, response time, and driving mode bias in a reward function, which balances and adjusts the learning rate and reward weight of the control strategy network.

8. The dynamic control method of a heat-dissipation drum brake based on reinforcement learning according to claim 1, characterized in that, The S5 specifically includes: S51, acquiring real-time thermal state data and inputting the thermal state data into an optimized thermal state prediction model, and predicting the temperature rise trend of the drum brake in a short time in the future by using the thermal state prediction model, for judging the temperature change trend and control demand; S52, inputting the thermal state data at the current time, the predicted temperature rise trend result, and the current driving mode state into the optimized control strategy network; S53, the control strategy network outputs a heat dissipation control strategy and control parameters according to the input information, the control strategy including heat dissipation mode selection, and the control parameters including a start signal of a heat dissipation device, a wind speed level, and a cooling medium flow rate; S54, generating a control instruction according to the control parameters output by the control strategy network, and issuing the control instruction to a corresponding heat dissipation execution mechanism through a communication interface, the heat dissipation execution mechanism including a fan, a heat dissipation air duct, or a liquid cooling device; S55, after receiving the control instruction, the heat dissipation execution mechanism executes the heat dissipation operation according to the specified parameters, including starting the heat dissipation device, adjusting the wind speed and volume, and regulating the flow rate of the heat dissipation liquid; S56, after completing the control task, the heat dissipation execution mechanism transmits execution state information and real-time temperature feedback data to the system.

9. A dynamic control system of a heat-dissipating drum brake based on reinforcement learning, based on the dynamic control method of a heat-dissipating drum brake based on reinforcement learning in any one of claims 1 to 8, characterized in that, The system includes the following modules: A data acquisition and preprocessing module for acquiring real-time operation data of the drum brake and preprocessing the operation data; A temperature rise prediction module for processing the input thermal state data and predicting the temperature change trend of the drum brake in a future time period; A control strategy generation module for outputting a heat dissipation control strategy and control parameters according to the thermal state data, the temperature rise trend, and the driving mode state; A heat dissipation execution module for receiving the control instruction issued by the control strategy generation module and controlling the operation state of the heat dissipation execution mechanism; A feedback and update module for receiving the execution state and temperature feedback information returned by the heat dissipation execution mechanism, and dynamically updating the parameters of the thermal state prediction model and the control strategy network in combination with the prediction error; A parameter optimization module for jointly optimizing the key hyperparameters in the temperature rise prediction module and the control strategy generation module.

Images