A method for detecting dynamic braking force of a drum

Abstract

The application provides a kind of detection method of dynamic braking force of drum, comprising: when detecting that vehicle enters drum area, drum real-time speed is collected, and non-contact optical field sensing system is started, for synchronously collecting total three-dimensional deformation time series data, temperature change time series data of drum area;Temperature change time series data and real-time speed are input into the coupling system of drum and wheel based on the finite element established in advance respectively, and the deformation time series data caused by temperature change and the deformation time series data caused by drum rotation centrifugal force are obtained;According to total three-dimensional deformation time series data of drum area, deformation time series data caused by temperature change and deformation time series data caused by drum rotation centrifugal force, deformation time series data caused only by braking force is determined;According to the deformation time series data caused only by braking force, dynamic braking force is estimated. By implementing the application, the accuracy and anti-interference ability of dynamic braking force detection can be improved.

Description

Technical Field

[0001] This invention belongs to the field of braking force detection technology, specifically relating to a method for detecting the dynamic braking force of a drum. Background Technology

[0002] In the field of dynamic braking force detection of rollers, traditional methods mostly rely on contact sensors (such as strain gauges and torque sensors) to obtain relevant data. However, these methods have significant limitations: Firstly, during braking, the rollers experience severe vibration and temperature increases. Contact sensors, directly mounted on the rollers or transmission structures, are susceptible to vibration interference, leading to signal distortion. Furthermore, high temperatures affect the stability and measurement accuracy of the sensors, especially under prolonged or high-intensity braking scenarios, where error accumulation becomes a prominent issue. Secondly, slippage may occur between the tire and the roller during braking. Torque values ​​calculated based on rotational speed cannot accurately reflect the true braking force. Additionally, torque sensors need to measure the response of the entire transmission chain, resulting in mechanical delays and an inability to precisely capture dynamic changes in braking force. Therefore, existing technologies suffer from insufficient accuracy and reliability, making it difficult to meet the high-precision detection requirements of complex braking scenarios. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide a method for detecting the dynamic braking force of a drum, so as to meet the high-precision detection requirements in complex braking scenarios.

[0004] To achieve the above objectives, the present invention provides the following technical solution: According to a first aspect, the present invention provides a method for detecting the dynamic braking force of a roller, comprising: when a vehicle is detected entering the roller region, acquiring the real-time rotational speed of the roller, and activating a non-contact optical field sensing system to simultaneously acquire total three-dimensional deformation time-series data and temperature change time-series data of the roller region; inputting the temperature change time-series data and the real-time rotational speed into a coupling system between the roller and the wheel pre-established based on finite element method, respectively, to obtain deformation time-series data caused by temperature change and deformation time-series data caused by the centrifugal force of roller rotation; determining deformation time-series data caused solely by braking force based on the total three-dimensional deformation time-series data of the roller region, the deformation time-series data caused by temperature change, and the deformation time-series data caused solely by the centrifugal force of roller rotation; and estimating the dynamic braking force based on the deformation time-series data caused solely by braking force.

[0005] According to a second aspect, the present invention provides a device for detecting the dynamic braking force of a roller, comprising: a data acquisition module, configured to acquire the real-time rotational speed of the roller when a vehicle is detected entering the roller area, and to activate a non-contact optical field sensing system for simultaneously acquiring total three-dimensional deformation time-series data and temperature change time-series data of the roller area; a first deformation time-series data determination module, configured to input the temperature change time-series data and the real-time rotational speed into a coupling system between the roller and the wheel pre-established based on finite element method, to obtain deformation time-series data caused by temperature change and deformation time-series data caused by the centrifugal force of roller rotation; a second deformation time-series data determination module, configured to determine deformation time-series data caused only by braking force based on the total three-dimensional deformation time-series data of the roller area, the deformation time-series data caused by temperature change, and the deformation time-series data caused by the centrifugal force of roller rotation; and a braking force estimation module, configured to estimate the dynamic braking force based on the deformation time-series data caused only by braking force.

[0006] According to a third aspect, an embodiment of the present invention provides an electronic device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of the method for detecting the dynamic braking force of a drum as described in the first aspect or any embodiment of the first aspect.

[0007] According to a fourth aspect, embodiments of the present invention provide a computer storage medium storing computer instructions thereon, which, when executed by a processor, implement the steps of the method for detecting the dynamic braking force of a drum as described in the first aspect or any embodiment of the first aspect.

[0008] This invention provides a method for detecting the dynamic braking force of a roller. A non-contact optical field sensing system synchronously collects the total three-dimensional deformation time-series data and temperature change time-series data of the roller area. Combined with a finite element method-based model of the roller-wheel coupling system, the deformation caused by temperature, centrifugal force, and braking force is separated. Finally, the dynamic braking force is estimated based on the deformation caused by pure braking force. The method proposed in this embodiment measures the physical deformation of the roller after being subjected to braking force (such as radial compression and circumferential torsion). This deformation has a direct mechanical relationship with the braking force and is unaffected by speed fluctuations caused by tire slippage. Furthermore, it overcomes the limitations of traditional contact sensors, which are susceptible to vibration and temperature rise interference. By using non-contact measurement and multi-factor deformation separation, it directly obtains core deformation data related to braking force, improving the accuracy and anti-interference capability of dynamic braking force detection, making it particularly suitable for complex braking scenarios.

[0009] Other advantages, objectives, and features of the invention will be set forth in the following description and will be apparent to those skilled in the art in some respects, or may be learned by practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0010] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following figures are provided for illustration: Figure 1 This is a flowchart illustrating a specific example of a method for detecting the dynamic braking force of a drum in this invention. Figure 2 This is a flowchart illustrating a specific example of estimating dynamic braking force based on time-series data of deformation caused solely by braking force and a second model in this invention. Figure 3 This is a flowchart illustrating a specific example of determining the deformation timing data caused solely by braking force in this invention. Figure 4 This is a schematic block diagram of a specific example of an electronic device in an embodiment of the present invention. Detailed Implementation

[0011] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can also refer to the internal connection of two components; and they can refer to a wireless connection or a wired connection. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0013] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0014] This invention provides a method for detecting the dynamic braking force of a drum, comprising: S101, when a vehicle is detected entering the roller area, the real-time rotation speed of the roller is collected, and the non-contact light field sensing system is activated to synchronously collect the total three-dimensional deformation time series data and temperature change time series data of the roller area. S102, input the temperature change time series data and real-time rotation speed into the coupling system of the drum and wheel based on the finite element method, respectively, to obtain the deformation time series data caused by temperature change and the deformation time series data caused by the centrifugal force of drum rotation; S103, based on the total three-dimensional deformation time series data of the drum area, the deformation time series data caused by temperature changes, and the deformation time series data caused by the centrifugal force of drum rotation, determine the deformation time series data caused only by braking force; S104. Estimate the dynamic braking force based on the deformation time series data caused solely by the braking force.

[0015] For example, when a vehicle is detected entering the roller area, the system initiates a data acquisition process. Specifically, the rotational speed of the roller is acquired in real time using devices such as a speed sensor. A non-contact optical field sensing system is then activated to synchronously acquire three-dimensional spatial deformation data of the roller area at different time points, forming a time-series sequence. Similarly, the non-contact optical field sensing system is used to acquire temperature change data of the roller area over time, forming a temperature time-series sequence. The non-contact optical field sensing system includes a structured light scanner, LiDAR, temperature acquisition devices (such as an infrared thermal imager), etc., and also includes a data synchronization and processing unit to ensure the synchronization of the acquired total three-dimensional deformation time-series data and the temperature change time-series data, and to perform preliminary processing on the acquired data.

[0016] Before implementing this scheme, a coupling system for the roller and wheel needs to be established in advance. The specific establishment method is as follows: First, manually record the material parameters, geometric parameters, and operating parameters of the roller and wheel. Material parameters include the elastic modulus, Poisson's ratio, density, etc., of the roller's material; the Shore hardness of the wheel's rubber and the elastic parameters of the cord layer, etc. Geometric parameters include the roller's diameter, length, and wall thickness; the wheel's specific shape encompassing the hub, rim, and tire; and the tire's tread pattern and carcass structure, etc. Operating parameters include different road surface adhesion coefficients and different braking pressures, etc. To avoid numerical simulation errors caused by aging and individual differences in roller material parameters, this embodiment further proposes establishing an aging level classification standard (e.g., brand new, slightly aged, heavily aged) based on historical data such as the roller's service life and cumulative braking count, and pre-setting corresponding material parameter correction coefficients for different levels. For example, the elastic modulus of a heavily aged roller can be corrected to 85% of the initial value, and the Poisson's ratio can be adjusted to 1.05 times the initial value, reducing the impact of individual differences on the simulation results.

[0017] After obtaining the above data, based on the material parameters, geometric parameters, and material parameter correction coefficients, adaptive meshing technology was used to refine the mesh of the roller surface and the wheel contact area. The tire portion was meshed using hyperelastic elements. The contact type between the roller and wheel was set to frictional contact, with the contact pair including the outer surface of the roller and the tire tread. An initial contact pressure distribution model was preset, such as a uniform distribution or a non-uniform distribution calculated theoretically from tire contact marks. Next, boundary conditions for the roller and wheel were set, including the six degrees of freedom at the roller shaft end, the rotational drive angular velocity, and the vertical load and braking torque of the wheel, etc.

[0018] Next, a force coupling is constructed based on mechanics, considering the mechanical interaction between the roller and the wheel, including normal contact force and tangential friction force. The normal contact stress is calculated according to Hertzian contact theory, and the tangential force follows Coulomb's friction law. The centrifugal force generated by the roller rotation is considered and applied as a volume load on the roller. The magnitude of the centrifugal force is calculated based on the roller speed and mass distribution. Then, a heat conduction model is established, setting the thermophysical properties of the material, such as thermal conductivity and specific heat capacity. At the same time, heat transfer boundary conditions are defined in the software, including convective heat transfer with air and thermal radiation. The influence of frictional heat generation during braking on the temperature field of the roller and wheel is also considered. Finally, the thermal expansion effect of the material caused by temperature change is considered, and thermal strain is introduced into the structural analysis as a volume load. The coefficient of thermal expansion is set according to the material parameters, establishing a two-way coupling between the temperature field and the deformation field. That is, temperature change leads to deformation, and deformation affects the temperature distribution. The coupling between the two is achieved through iterative calculation.

[0019] After constructing the geometric model, setting the force coupling rules and boundary conditions as described above, a roller braking experiment was conducted on the model to test data such as deformation, temperature, and braking force under different working conditions. The calculation results of the finite element model were compared with the experimental data. If the deviation between the calculation results and the experimental data was large, the cause was analyzed and the model was corrected, such as adjusting material parameters, optimizing mesh generation, and adjusting contact settings, until the deviation between the calculation results and the experimental data was less than the acceptable error range, such as 3%. Finally, the verified and optimized finite element model was encapsulated into a callable function or module, providing input and output interfaces for easy integration into the detection method. During encapsulation, the key parameter interfaces of the model, such as roller speed and temperature change time series data, were retained to allow for real-time input of actual detection data.

[0020] The coupling system of the roller and wheel constructed based on the above method inputs temperature change time-series data and real-time rotational speed into the system. Based on temperature changes, the time-series sequence of roller deformation caused by temperature effects is calculated. Based on the real-time rotational speed, the time-series data of deformation caused by the centrifugal force generated by roller rotation is calculated. The total three-dimensional deformation time-series data of the roller region is obtained. This data includes the comprehensive deformation of the roller caused by all factors during braking. By subtracting the deformation time-series data caused by temperature changes and the deformation time-series data caused by the centrifugal force of roller rotation from the total three-dimensional deformation time-series data, the influence of temperature and centrifugal force factors is eliminated, thus obtaining the deformation time-series data caused only by braking force.

[0021] Finally, a pre-trained neural network model, such as a convolutional model or a deep learning model, is used as input, taking the deformation time series data caused solely by braking force as input, and outputting the estimated dynamic braking force.

[0022] This invention provides a method for detecting the dynamic braking force of a roller. A non-contact optical field sensing system synchronously collects the total three-dimensional deformation time-series data and temperature change time-series data of the roller area. Combined with a finite element method-based model of the roller-wheel coupling system, the deformation caused by temperature, centrifugal force, and braking force is separated. Finally, the dynamic braking force is estimated based on the deformation caused by pure braking force. The method proposed in this embodiment measures the physical deformation of the roller after being subjected to braking force (such as radial compression and circumferential torsion). This deformation has a direct mechanical relationship with the braking force and is unaffected by speed fluctuations caused by tire slippage. Furthermore, it overcomes the limitations of traditional contact sensors, which are susceptible to vibration and temperature rise interference. By using non-contact measurement and multi-factor deformation separation, it directly obtains core deformation data related to braking force, improving the accuracy and anti-interference capability of dynamic braking force detection, making it particularly suitable for complex braking scenarios.

[0023] As an optional implementation, before determining the deformation time series data caused solely by braking force based on the total three-dimensional deformation time series data of the roller region, the deformation time series data caused by temperature changes, and the deformation time series data caused by the centrifugal force of the roller rotation, the process includes: calling a pre-deployed accelerometer to collect the rolling rotation acceleration; determining the deformation-estimated acceleration based on the total three-dimensional deformation time series data; determining the correlation coefficient between the rolling rotation acceleration and the deformation-estimated acceleration; if the time domain correlation coefficient is greater than or equal to a preset threshold, the total three-dimensional deformation time series data is deemed valid.

[0024] For example, before calculating the deformation time series data caused solely by braking force, the rolling acceleration collected by pre-deployed accelerometers is invoked. These accelerometers are symmetrically deployed in the non-contact area at the end of the roller shaft. When the vehicle's wheel axle passes over the photoelectric sensor at the roller inlet, the acquisition clocks of the accelerometers and the light field sensing system are simultaneously triggered, ensuring data timestamp consistency. Then, based on the total three-dimensional deformation time series data, the deformation-estimated acceleration is determined. Specifically, the total three-dimensional deformation time series data output by the light field sensing system is a point cloud sequence. Each point cloud contains circumferential angle, axial position, radial deformation, and a timestamp. Cubic spline interpolation is used to spatiotemporally reconstruct the discrete point clouds, constructing a continuous function. , Let z represent the circumferential angle, z represent the axial position, and t represent time, ensuring the continuity of deformation data in both time and space dimensions. Then, the radial deformation represented by the continuous function is treated as the displacement of particles on the roller surface, and the acceleration is calculated using the second-order central difference method. The specific formula is as follows: Finally, a weighted average of the deformation accelerations over the 360-degree circumferential direction and the entire axial length is calculated. The weighting coefficients are pre-calibrated based on the stress distribution at each point on the roller surface, such as a weight of 0.8 for the contact area and 0.2 for the non-contact area. After obtaining the estimated deformation acceleration, the correlation coefficient between it and the rolling rotation acceleration is determined using the Pearson correlation coefficient formula. When the correlation coefficient is greater than a preset threshold, such as 0.9, it indicates that the two are relatively close with a small error, and therefore, the total three-dimensional deformation time series data can be determined to be valid.

[0025] This invention provides a method for detecting the dynamic braking force of a roller. By calling a pre-deployed accelerometer to collect the rolling rotation acceleration, and determining the deformation-induced acceleration based on the total three-dimensional deformation time series data, the validity of the total three-dimensional deformation time series data is determined by calculating the time domain correlation coefficient between the two. Through the above verification mechanism, invalid data caused by sensor failure, external interference, and other factors can be effectively eliminated, ensuring that the data used for subsequent braking force calculation is true and reliable, thereby improving the accuracy of dynamic braking force detection.

[0026] As an optional implementation, a method for detecting the dynamic braking force of a drum further includes: if the time-domain correlation coefficient is less than a preset threshold, then the total three-dimensional deformation time-series data is determined to be invalid, and the following steps are performed: The database is accessed, and the comprehensive difference between the current operating condition and historical data is calculated in priority order of vehicle type, drum speed, and ambient temperature. The set of historical total 3D deformation time series data with the smallest comprehensive difference and the most recent timestamp is selected. It is determined whether the difference between the drum speed currently detected and the speed corresponding to the historical total 3D deformation time series data exceeds a preset threshold. If it exceeds the threshold, the historical total 3D deformation time series data is linearly interpolated and scaled to synchronize the time axis with the current detection, and the time-synchronized historical total 3D deformation time series data is used as the total 3D deformation time series data. If it does not exceed the threshold, the historical total 3D deformation time series data is used as the total 3D deformation time series data.

[0027] For example, when the total three-dimensional deformation time series data is determined to be invalid, the system automatically triggers a historical data compensation mechanism. Specifically, it first calls the database and calculates the Euclidean distance between the current operating condition and the historical data according to the priority order of vehicle type, roller speed, and ambient temperature. Vehicle type is matched first; if the current vehicle is a sedan, all historical data of sedans are filtered from the database. Next, roller speed is matched, and the difference between the currently detected roller speed and the roller speed in the historical data is calculated. Finally, ambient temperature is matched, and the difference between the current ambient temperature and the ambient temperature in the historical data is calculated. Then, through a weighted method, the comprehensive difference of these three dimensions is obtained, and the set of historical total three-dimensional deformation time series data with the smallest comprehensive difference and the most recent timestamp is selected to ensure the highest matching degree between the historical data and the current operating condition.

[0028] In this embodiment, minimizing the drum speed during the initial screening only ensures optimal static parameter matching between historical data and current operating conditions, but it cannot directly reflect the consistency between historical data and current detection in the dynamic time dimension. Therefore, for the set of historical total 3D deformation time series data with the smallest comprehensive difference and the most recent timestamp, it is necessary to further determine the impact of its speed difference on the time axis. If the speed difference between the current detection speed and the speed corresponding to the historical total 3D deformation time series data is too large, even if the comprehensive difference is small, it may lead to the historical total 3D deformation time series data being out of sync with the current detection. To solve this problem, this embodiment calculates the difference between the currently detected drum speed and the speed corresponding to the historical total 3D deformation time series data. When the speed difference does not exceed a preset threshold, it indicates that the difference is small and the impact on the time series is small, which can be ignored. In this case, the historical total 3D deformation time series data can be used as the total 3D deformation time series data. When the speed difference exceeds the threshold, it indicates that the difference on the time series is large, and a linear interpolation scaling method is used to calibrate the time axis. The historical total 3D deformation time series data after time synchronization calibration is used as the total 3D deformation time series data.

[0029] This invention provides a method for detecting the dynamic braking force of a drum. When the total three-dimensional deformation data is invalid, the method retrieves the historical data from the database that has the highest matching degree with the current operating conditions (vehicle type, speed, temperature), and achieves time axis synchronization through speed difference calibration. This mechanism ensures the continuity of detection, avoids detection interruptions caused by real-time data failure, and improves detection efficiency. Simultaneously, the priority matching and speed calibration strategy ensures the dynamic consistency between historical data and the current operating conditions. Linear interpolation scaling can solve the problem of time axis asynchrony caused by excessive speed differences, minimizing the impact of missing data on the results.

[0030] As an optional implementation, the total three-dimensional deformation time-series data includes data source labels to characterize whether the data is valid data or data compensated using historical data. Based on the deformation time-series data caused solely by braking force, the dynamic braking force is estimated, including: When the total three-dimensional deformation time series data is valid, the deformation time series data caused only by braking force is input into the first model to estimate the dynamic braking force; when the total three-dimensional deformation time series data is data compensated by historical data, the deformation time series data caused only by braking force is input into the second model to estimate the dynamic braking force.

[0031] As an optional implementation, the deformation time-series data caused solely by braking force is input into the first model to estimate the dynamic braking force. This includes: extracting spatiotemporal features from the deformation time-series data caused solely by braking force to obtain a feature matrix containing spatiotemporal features; inputting the feature matrix into a convolutional neural network (Long Short-Term Memory network) to capture bidirectional dependencies in the time-series data using forward and backward hidden states, generating a time-series feature vector; weighting the time-series feature vector through an attention mechanism to enhance key time-point features strongly correlated with braking force; and outputting the dynamic braking force estimate by a fully connected layer.

[0032] For example, the first model can be a pre-trained Convolutional Neural Network-Long Short-Term Memory Network (CNN-LSTM) hybrid model. The training process of the CNN-LSTM hybrid model includes: First, constructing a dataset and collecting effective data from vehicle braking experiments, including processed deformation time-series data caused solely by braking force and synchronously recorded real dynamic braking force values. The real dynamic braking force values ​​can be measured by a wheel torque measurement system. Specifically, in the vehicle braking experiment, a torque sensor is integrated between the wheel hub and the drive shaft, ensuring that the sensor is concentric with the wheel's rotation axis to reduce the impact of mechanical deviations on the measurement. When the vehicle enters the roller area and performs braking, the torque sensor senses the torque signal transmitted during wheel braking in real time. This signal is converted into an electrical signal by the signal acquisition module and then timestamped by the data synchronization unit with the roller speed and deformation time-series data. Then, the acquired torque signal is filtered to remove high-frequency noise generated by braking vibration. Based on the mechanical relationship: Braking Force = Torque / Effective Wheel Radius, the filtered torque value is converted into dynamic braking force data, which serves as the real braking force label required for model training.

[0033] Then, the deformation data is subjected to short-time Fourier transform to generate a time-frequency matrix, and statistical features such as peak value and root mean square value are calculated to construct a spatiotemporal feature matrix. The dataset is then divided into training, validation, and test sets. During training, mean square error is used as the loss function, the Adam optimizer is used, a maximum of 100 iterations are set, an early stopping strategy is adopted, and L2 regularization is added to the fully connected layers. Finally, during training, the weights are randomly initialized, the feature matrix of the training set is input into the CNN to extract spatial features, and then fed into a bidirectional LSTM to capture temporal dependencies. Key time point features are weighted through an attention mechanism, the fully connected layers output predicted values, the loss is calculated, and the weights are updated in reverse using the Adam optimizer. After each training round, the validation set is used to evaluate and adjust the hyperparameters. After training is completed, the generalization ability is evaluated on the test set, and the model with the smallest loss on the validation set is saved for dynamic braking force estimation.

[0034] A short-time Fourier transform (SFT) is performed on the deformation time-series data caused solely by braking force to generate a time-frequency matrix. Simultaneously, statistical characteristics such as peak value and root mean square value in the time domain are calculated, resulting in a multidimensional spatiotemporal feature matrix containing both time-domain and frequency-domain features. The Hanning window is chosen as the window function for the SFT, with its window length set to 1.5 times the typical time constant of the braking process to balance time and frequency resolution.

[0035] The multidimensional feature matrix is ​​input into a CNN-LSTM hybrid model, which includes convolutional neural network layers and long short-term memory (LSTM) network layers. The convolutional neural network layers consist of three one-dimensional convolutional layers with 64, 128, and 256 kernels respectively, all with a kernel size of 5, a stride of 1, and padding of 'same'. Batch Normalization and ReLU activation functions are added after each layer to extract local spatial and abstract features from the multidimensional feature vector. The LSTM network layers consist of two bidirectional LSTM layers with 128 units each, used to capture bidirectional dependencies in temporal data through forward and backward hidden states, with particular attention to features at key time points such as the start, peak, and end of the braking process.

[0036] An attention mechanism is applied after the output of the LSTM layer to calculate the attention weight for each time step, thereby enhancing the features of key time points strongly correlated with braking force. The weighting process of the temporal feature vector sequence in the attention mechanism is as follows: First, the feature vector sequence output by the bidirectional LSTM is mapped to an intermediate vector through a learnable weight matrix and bias mapping. After tanh activation, it is multiplied by the learnable attention vector u to obtain the attention score for each time step. The attention score is then normalized using the softmax function to obtain the weight sequence. Key time points strongly correlated with braking force (such as the braking peak moment and the moment of sudden deformation) are assigned higher weights due to their more significant features. The 256-dimensional feature vector of each time step is multiplied by its corresponding weight and then summed to obtain the weighted 256-dimensional key temporal feature vector. Finally, dimensionality reduction is performed through a fully connected layer to output the estimated value of dynamic braking force.

[0037] This invention provides a method for detecting the dynamic braking force of a drum. The first model extracts spatiotemporal features, captures temporal dependencies using a bidirectional LSTM, and strengthens key features through an attention mechanism to output an estimated braking force. This model can deeply mine the spatiotemporal correlation of deformation data. Short-time Fourier transform and convolutional layers can analyze the spatial distribution and frequency characteristics of deformation. The bidirectional LSTM can capture the temporal dependencies during the braking process. The attention mechanism weights key time points such as braking peaks, solving the problem of key information being overwhelmed due to the equal processing of all time points by traditional models, and improving the ability to capture the changing trend of dynamic braking force.

[0038] As an optional implementation, the dynamic braking force is estimated based on the deformation time-series data caused solely by the braking force and a second model, such as... Figure 2 As shown, it includes: S201, determine the multidimensional difference coefficient based on current operating conditions and historical data; S202, when the multidimensional difference coefficient exceeds the preset threshold, a spatial attention weight matrix is ​​generated according to the preset function, and pixel-by-pixel weighted suppression is performed on the area in the deformation time series data caused only by braking force that does not match the current working condition. S203 uses bidirectional gated recurrent units to capture the temporal features of deformation in weighted and suppressed deformation time series data, and uses convolutional kernels to extract spatial distribution features; S204 concatenates the temporal dimension features and spatial distribution features to construct the input vector; S205, Based on the deformation time series data caused solely by braking force, determine the variance of the deformation data for this instance; S206, Obtain the average variance of deformation data in historical data; S207. Determine the fluctuation attenuation factor based on the variance of the deformation data and the average variance of the deformation data. S208, constructing an exponential decay factor based on multidimensional difference coefficients; S209, adjust the input vector according to the fluctuation decay factor and the exponential decay factor to obtain the weighted input vector; S210, the weighted input vector is input into the pre-trained gradient boosting decision model to obtain the estimate of the dynamic braking force.

[0039] For example, the multidimensional difference coefficient is determined according to the following formula. : ; in, Indicates the current operating condition. j Item parameters, Indicates the first in historical data j Item parameters, Indicates the first j The importance weight of each parameter, such as the importance weight of parameters corresponding to speed, axle load, and temperature. The values ​​are 0.4, 0.3, and 0.3 respectively. For the first j The sensitivity coefficients of the parameters are, for example, 0.1, 0.05 and 0.02 respectively, but this embodiment does not limit them.

[0040] when If the value exceeds a preset threshold, it indicates a significant difference between the current operating condition and historical data. Therefore, it is necessary to suppress the mismatched regions. The preset threshold can be 0.6, but this embodiment does not limit it. In this embodiment, a spatial attention weight matrix is ​​used to perform pixel-by-pixel weighted suppression on regions where the deformation time-series data caused solely by braking force does not match the current operating condition. The spatial attention weight matrix is ​​determined according to a preset function. Specifically, firstly, the preset function is used: Generate weight matrix ,in, for sigmoid function, The gradient feature of the roller surface position reflects the spatial rate of change of the braking force distribution. The preset function can also be other function forms, as long as the higher the matching degree, the higher the output weight. Next, pixel-by-pixel weighted suppression is applied to regions that do not match the current operating condition. Specifically, regions that do not match the current operating condition are... Then the weighted suppression process is as follows: = 。

[0041] The suppressed deformation data is unfolded into a two-dimensional matrix over time steps. A sliding convolution with a 3×3 kernel is used to extract local deformation patterns. Then, a 2×2 max-pooling layer is applied to reduce the feature dimensionality, retaining the most significant spatial features and outputting a 64-dimensional spatial feature vector. Simultaneously, the spatial feature vector from each time step is input into a bidirectional gated recurrent unit (GRU). The feedforward layer captures past temporal dependencies, and the backward layer captures future temporal dependencies. Each GRU has 64 neurons, ultimately outputting a 64-dimensional temporal feature vector. The 64-dimensional spatial feature vector and the 64-dimensional temporal feature vector are concatenated column-wise to form a 128-dimensional input vector.

[0042] To mitigate the impact of mismatches with current operating conditions or excessive data fluctuations, this embodiment employs a dual-weight attenuation method to simultaneously suppress the effects of both issues. Regarding the problem of operating condition data mismatch, based on the aforementioned findings, a multidimensional difference coefficient has been obtained. An exponential attenuation factor can be constructed based on this coefficient. Specifically, to achieve the effect that the greater the difference in operating conditions, the stronger the suppression of the input vector, this embodiment uses the following formula to obtain the exponential attenuation factor. : ; in, The attenuation coefficient can be set to 2 based on experience, but this embodiment does not limit it.

[0043] Regarding the impact of data fluctuations, this embodiment obtains the average variance of historical data under the same working conditions from the historical database, calculates the average variance of the current deformation data, and then uses the variance of the current deformation data as a basis for further analysis. and the average variance of deformation data Determine the fluctuation attenuation factor , .

[0044] Adjust the input vector based on the fluctuation decay factor and the exponential decay factor. The weighted input vector is obtained. Specifically: .Will The input is fed into a pre-trained gradient boosting decision model to obtain an estimate of the dynamic braking force.

[0045] The gradient boosting decision model is trained as follows: It uses reduced-dimensional deformation features and corresponding operating parameters from historical valid data as input, and corresponding braking force as output. The gradient boosting algorithm iteratively constructs multiple decision trees, with each tree learning the residual of the previous tree. The number of trees is set to 100, the maximum depth of a single tree is 5, and the learning rate is 0.1 to balance model complexity and generalization ability. The method for obtaining the true braking force labels is as described above and will not be elaborated here. The weights of the loss function are dynamically adjusted based on the multidimensional difference coefficient. The loss function formula is: ； in, Mean square error, The mean absolute error, This is the divergence scaling factor, which can be 0.5. for KL Divergence, representing a measure of the current data distribution. p Distribution of historical data q The difference, when When the difference is large, the model is forced to learn the distribution difference.

[0046] This invention provides a method for detecting the dynamic braking force of a drum. It suppresses mismatch regions through a multidimensional difference coefficient, extracts spatiotemporal features using a bidirectional gated recurrent unit and convolutional kernel, and introduces fluctuation decay and exponential decay factors to optimize the input. This model addresses the difference between historical compensation data and current operating conditions by suppressing mismatch regions through a spatial attention weight matrix, reducing the impact of data bias. The bidirectional gated recurrent unit and convolutional kernel capture temporal trends and spatial patterns respectively, fully utilizing the multidimensional information of deformation data. The fluctuation decay factor weakens data jitter, and the exponential decay factor reduces the weight of data with large operating condition differences. This dual optimization improves the model's tolerance to data bias, ensuring the reliability of detection under compensated data.

[0047] As an optional implementation, based on the total three-dimensional deformation time-series data of the drum region, the deformation time-series data caused by temperature changes, and the deformation time-series data caused by the centrifugal force of drum rotation, the deformation time-series data caused solely by braking force is determined, including: The total three-dimensional deformation time series data, the deformation time series data caused by temperature change, and the deformation time series data caused by the centrifugal force of the drum rotation were sampled according to the target interval time to obtain the total three-dimensional deformation dataset, the temperature deformation dataset, and the centrifugal force deformation dataset, respectively. Total 3D deformation data, temperature deformation data, and centrifugal force deformation data collected at the target time slot were extracted from the three deformation datasets respectively. For the three types of deformation data, such as Figure 3 As shown, perform the following steps respectively: S1, perform secondary sampling in the circumferential angle. In the secondary sampling, uniform sampling is performed in the circumferential angle to obtain multiple circumferential sampling angles. S2, perform secondary sampling in the axial direction. In the secondary sampling, in the axial direction, according to the pre-calibrated segment intervals and segment types, different sampling densities are determined in different segment intervals in the axial direction. Based on the sampling density, the number of multiple axial sampling points under different segments is obtained. S3, taking the target starting point as the origin, obtains multiple sampling point data based on multiple circumferential sampling angles and the number of multiple axial sampling points under different segments; S4, subtract the data of each target sampling point corresponding to the temperature deformation dataset and the data of each target sampling point corresponding to the centrifugal force deformation dataset from the data of each target sampling point corresponding to the total three-dimensional deformation dataset collected in the same time slot, to obtain the deformation data of each target sampling point caused only by the braking force under the target time slot. S5, perform principal component analysis on the deformation data of each target sampling point under the target time slot caused only by braking force to obtain the reduced dimension deformation data of the target time slot caused only by braking force; For the total three-dimensional deformation data, temperature deformation data and centrifugal force deformation data collected at different time slots, repeat steps S1-S5 to obtain the deformation data caused only by braking force in different time slots with reduced dimensions. Based on the deformation data caused solely by braking force in different time slots with reduced dimensions, time-series data of deformation caused solely by braking force are obtained.

[0048] For example, according to a preset target interval, such as every 10 milliseconds, the total three-dimensional deformation time series data, the deformation time series data caused by temperature changes, and the deformation time series data caused by the centrifugal force of the rotating drum are synchronously sampled, thereby generating the total three-dimensional deformation dataset, the temperature deformation dataset, and the centrifugal force deformation dataset, respectively. These datasets all contain deformation information at different time points.

[0049] Extract the total 3D deformation data, temperature deformation data, and centrifugal force deformation data within the same target time slot from these three datasets, and then perform the following operations on these three types of data in sequence: In the secondary sampling of circumferential angles, since the roller has a circular structure with a circumferential angle range of 0-360 degrees, following the principle of uniform sampling, if sampling is set to occur every 10 degrees, 36 circumferential sampling angles will be obtained (360 ÷ 10 = 36), covering different positions along the entire circumference of the roller. In the secondary sampling of axial direction, the axial direction is pre-divided into different segments based on the structural characteristics and stress conditions of the roller, such as left, middle, and right segments. The type of each segment is also determined, such as a segment of primary concern or a segment of general concern. A higher sampling density is set for the middle segment, which is of primary concern, while the left and right segments, as segments of general concern, have a lower sampling density. Based on this sampling density, the number of axial sampling points corresponding to different segments can be obtained. For example, a middle segment with a length of 50 cm will have 100 sampling points, and a left and right segment of 30 cm each will have 30 sampling points.

[0050] Using a fixed point along the axis of the roller, such as the starting point at the left end, as the target starting point and regarded as the origin, and combining the previously obtained multiple circumferential sampling angles and the number of multiple axial sampling points under different segments, the specific position of each sampling point in three-dimensional space is determined, and then the data of multiple sampling points corresponding to these positions are obtained. These data reflect the deformation of the roller at different positions under this time slot.

[0051] For the same target time slot, the data of each target sampling point in the total three-dimensional deformation dataset is subtracted from the data of the corresponding sampling points in the temperature deformation dataset and the centrifugal force deformation dataset. By calculating the difference, the deformation data of each target sampling point in that time slot caused only by the braking force can be obtained. This is because the total deformation is caused by the combined action of the braking force, temperature change and centrifugal force. Subtracting the influence of the latter two separates the deformation caused by the braking force alone.

[0052] Then, principal component analysis is performed on all deformation data caused solely by braking force in this time slot. Principal component analysis transforms multiple related deformation data indicators into a few unrelated principal components, retaining most of the key information, thereby achieving data dimensionality reduction and obtaining the dimensionality-reduced deformation data caused solely by braking force for the target time slot. The process of principal component analysis is as follows: The braking force deformation data corresponding to all target sampling points in this time slot constitute a multidimensional dataset. The deformation data of each sampling point contains information from multiple dimensions. This multidimensional dataset is standardized to eliminate the influence of differences in dimensions or orders of magnitude. Typically, the mean of each dimension is subtracted from the data in that dimension, and then divided by the standard deviation, resulting in a mean of 0 and a standard deviation of 1. Next, the covariance matrix of the standardized dataset is calculated. The covariance matrix reflects the correlation between different dimensions. The elements in the matrix represent the covariance of corresponding two dimensions; positive values ​​indicate positive correlation, negative values ​​indicate negative correlation, and the absolute value reflects the degree of correlation. Then, the eigenvalues ​​and corresponding eigenvectors of the covariance matrix are solved. The eigenvalues ​​represent the contribution of the principal component corresponding to the eigenvector to the data variation; the larger the eigenvalue, the more original data information the corresponding principal component contains. The eigenvectors represent the direction of the principal components, i.e., the new dimensions formed by linear combination of the original dimensions.

[0053] Next, the eigenvectors are sorted according to the magnitude of the eigenvalues, arranged in descending order of eigenvalues, and the corresponding eigenvectors are also sorted in turn. The first k eigenvectors are selected as principal components. The determination of k is usually based on the cumulative contribution rate, that is, the proportion of the sum of the first k eigenvalues ​​to the total sum of all eigenvalues. When the cumulative contribution rate reaches a preset threshold (such as 85% or above), it is considered that these k principal components can retain most of the key information of the original data. At this time, k is the number of dimensions after dimensionality reduction.

[0054] Finally, the standardized original dataset is multiplied by the matrix composed of the selected first k feature vectors to obtain the dimensionality-reduced dataset. The dimension of each sample point in this dataset is reduced from the original multidimensionality to k dimensions, and each dimension corresponds to a principal component. The resulting dimensionality-reduced dataset is the deformation data caused only by braking force in the target time slot.

[0055] In the same manner, the steps from S1 to S5 were repeated for the total three-dimensional deformation data, temperature deformation data, and centrifugal force deformation data collected in different time slots to obtain the deformation data caused only by braking force after dimensional reduction in each time slot. Finally, the data from these different time slots were arranged in chronological order to form the deformation time series data caused only by braking force. This data completely reflects the change of the roller deformation over time under the action of braking force alone during the entire detection process.

[0056] This invention provides a method for detecting the dynamic braking force of a drum. Through layered sampling (circumferential uniform sampling + axial differential sampling) and principal component analysis, pure braking force deformation time-series data are separated and dimensionality-reduced from the total deformation. In layered sampling, circumferential uniform sampling covers the entire circumference of the drum, while axial differential sampling focuses on key areas such as the braking contact zone. Compared to full sampling, this reduces the amount of data significantly and lowers the computational load. Principal component analysis, while reducing dimensionality, retains key deformation patterns related to braking force (such as gradient distribution along the axial direction), avoiding model overfitting caused by high-dimensional data. This provides high-quality input for subsequent model calculations, ensuring the effectiveness and accuracy of the pure braking force deformation data.

[0057] This invention provides a device for detecting the dynamic braking force of a drum, comprising: The acquisition module is used to acquire the real-time rotation speed of the roller when a vehicle is detected entering the roller area, and to activate the non-contact light field sensing system to simultaneously acquire the total three-dimensional deformation time series data and temperature change time series data of the roller area. The first deformation time series data determination module is used to input the temperature change time series data and the real-time rotation speed into the coupling system of the drum and wheel based on the finite element method, respectively, to obtain the deformation time series data caused by temperature change and the deformation time series data caused by the centrifugal force of the drum rotation. The second deformation timing data determination module is used to determine the deformation timing data caused only by braking force based on the total three-dimensional deformation timing data of the drum area, the deformation timing data caused by temperature changes, and the deformation timing data caused by the centrifugal force of drum rotation. The braking force estimation module is used to estimate the dynamic braking force based on the deformation time series data caused solely by braking force.

[0058] This application also provides an electronic device, such as... Figure 4 As shown, processor 501 and memory 502 are connected via a bus or other means.

[0059] Processor 501 can be a central processing unit (CPU). Processor 501 can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips.

[0060] The memory 502, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the drum dynamic braking force detection method in this embodiment of the invention. The processor executes various functional applications and data processing by running the non-transitory software programs, instructions, and modules stored in the memory.

[0061] Memory 502 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the processor, etc. Furthermore, the memory may include high-speed random access memory and non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 502 may optionally include memory remotely located relative to the processor, which can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0062] The one or more modules are stored in the memory 502, and when executed by the processor 501, they perform actions such as... Figure 1 The method for detecting the dynamic braking force of the drum in the illustrated embodiment.

[0063] For specific details regarding the aforementioned electronic devices, please refer to the relevant documentation. Figure 1 The relevant descriptions and effects in the illustrated embodiments are for understanding purposes only and will not be repeated here.

[0064] This embodiment also provides a computer storage medium storing computer-executable instructions that can execute the method for detecting the dynamic braking force of the drum in any of the above-described method embodiments. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium may also include combinations of the above types of memory.

[0065] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made to it in form and detail without departing from the scope defined by the claims of the present invention.

Claims

1. A method for detecting the dynamic braking force of a drum, characterized in that, include: When a vehicle is detected entering the roller area, the real-time rotation speed of the roller is collected, and the non-contact light field sensing system is activated to synchronously collect the total three-dimensional deformation time series data and temperature change time series data of the roller area. The temperature change time series data and real-time rotation speed are respectively input into the coupling system of the drum and wheel based on the finite element method, to obtain the deformation time series data caused by temperature change and the deformation time series data caused by the centrifugal force of drum rotation. Based on the total three-dimensional deformation time series data of the drum area, the deformation time series data caused by temperature changes, and the deformation time series data caused by the centrifugal force of drum rotation, the deformation time series data caused only by braking force is determined. Estimate dynamic braking force based on time-series data of deformation caused solely by braking force.

2. The method for detecting the dynamic braking force of a drum according to claim 1, characterized in that, Based on the total three-dimensional deformation time series data of the drum region, the deformation time series data caused by temperature changes, and the deformation time series data caused by the centrifugal force of drum rotation, before determining the deformation time series data caused solely by braking force, the following steps are taken: The pre-deployed accelerometers are used to collect rolling and rotational acceleration. Based on the total three-dimensional deformation time series data, the deformation-estimated acceleration is determined; The correlation coefficient between the two is determined by calculating the acceleration based on the rolling acceleration and the deformation. If the time domain correlation coefficient is greater than or equal to the preset threshold, the total three-dimensional deformation time series data is deemed valid.

3. The method for detecting the dynamic braking force of a drum according to claim 2, characterized in that, Also includes: If the time-domain correlation coefficient is less than a preset threshold, the total three-dimensional deformation time-series data is deemed invalid, and the following steps are executed: The database is called to calculate the comprehensive difference between the current working condition and historical data in the priority order of vehicle type, drum speed and ambient temperature. The set of historical total three-dimensional deformation time series data with the smallest comprehensive difference and the most recent timestamp is selected. Determine whether the difference between the currently detected drum speed and the speed corresponding to the historical total three-dimensional deformation time series data exceeds a preset threshold; If the time exceeds the limit, linear interpolation scaling is performed on the historical total three-dimensional deformation time series data to synchronize the time axis with the current detection, and the time-synchronized historical total three-dimensional deformation time series data is used as the total three-dimensional deformation time series data. If the value does not exceed a certain threshold, then the historical total three-dimensional deformation time series data will be used as the total three-dimensional deformation time series data.

4. A method for detecting the dynamic braking force of a drum according to any one of claims 1-3, characterized in that, The total three-dimensional deformation time-series data includes data source labels to characterize whether the data is valid or compensated using historical data. Based on the deformation time-series data caused solely by braking force, the dynamic braking force is estimated, including: When the total three-dimensional deformation time series data is valid, the deformation time series data caused only by the braking force is input into the first model to estimate the dynamic braking force. When the total three-dimensional deformation time series data is the data compensated using historical data, the dynamic braking force is estimated based on the deformation time series data caused only by braking force and the second model.

5. The method for detecting the dynamic braking force of a drum according to claim 4, characterized in that, Input the deformation time-series data caused solely by braking force into the first model to estimate the dynamic braking force, including: Spatiotemporal features are extracted from time-series data of deformation caused solely by braking force to obtain a feature matrix containing spatiotemporal features; The feature matrix is ​​input into a bidirectional long short-term memory network, and the bidirectional dependencies in the time series data are captured by the forward and backward hidden states to generate time series feature vectors. By weighting the temporal feature vector through an attention mechanism, the key time point features that are strongly correlated with braking force are enhanced, and the dynamic braking force estimate is output by the fully connected layer.

6. The method for detecting the dynamic braking force of a drum according to claim 4, characterized in that, The second model is a gradient boosting decision model. Based on the time-series data of deformation caused solely by braking force and the second model, it estimates the dynamic braking force, including: Determine the multidimensional difference coefficient based on current operating conditions and historical data; When the multidimensional difference coefficient exceeds the preset threshold, a spatial attention weight matrix is ​​generated according to the preset function, and pixel-by-pixel weighted suppression is performed on the deformation time series data caused solely by braking force that does not match the current working condition. We use bidirectional gated recurrent units to capture the temporal features of deformation in weighted and suppressed deformation time series data, and use convolutional kernels to extract spatial distribution features. The temporal and spatial distribution features are concatenated to construct the input vector; Based on the deformation time series data caused solely by braking force, determine the variance of the deformation data for this study; Obtain the average variance of deformation data from historical data; The fluctuation attenuation factor is determined based on the variance and average variance of the deformation data. Construct an exponential decay factor based on the multidimensional difference coefficient; The input vector is adjusted based on the fluctuation decay factor and the exponential decay factor to obtain the weighted input vector. The weighted input vector is fed into a pre-trained gradient boosting decision model to obtain an estimate of the dynamic braking force.

7. The method for detecting the dynamic braking force of a drum according to claim 1, characterized in that, Based on the total three-dimensional deformation time series data of the drum region, the deformation time series data caused by temperature changes, and the deformation time series data caused by the centrifugal force of drum rotation, the deformation time series data caused solely by braking force is determined, including: The total three-dimensional deformation time series data, the deformation time series data caused by temperature change, and the deformation time series data caused by the centrifugal force of the drum rotation were sampled according to the target interval time to obtain the total three-dimensional deformation dataset, the temperature deformation dataset, and the centrifugal force deformation dataset, respectively. Extract the total 3D deformation data, temperature deformation data, and centrifugal force deformation data collected at the target time slot from the three deformation datasets respectively. Perform the following steps for each of the three deformation datasets: S1, perform secondary sampling in the circumferential angle. In the secondary sampling, uniform sampling is performed in the circumferential angle to obtain multiple circumferential sampling angles. S2, perform secondary sampling in the axial direction. In the secondary sampling, in the axial direction, according to the pre-calibrated segment intervals and segment types, different sampling densities are determined in different segment intervals in the axial direction. Based on the sampling density, the number of multiple axial sampling points under different segments is obtained. S3, taking the target starting point as the origin, obtains multiple sampling point data based on multiple circumferential sampling angles and the number of multiple axial sampling points under different segments; S4, subtract the data of each target sampling point corresponding to the temperature deformation dataset and the data of each target sampling point corresponding to the centrifugal force deformation dataset from the data of each target sampling point corresponding to the total three-dimensional deformation dataset collected in the same time slot, to obtain the deformation data of each target sampling point caused only by the braking force under the target time slot. S5, perform principal component analysis on the deformation data of each target sampling point under the target time slot caused only by braking force to obtain the reduced dimension deformation data of the target time slot caused only by braking force; For the total three-dimensional deformation data, temperature deformation data and centrifugal force deformation data collected at different time slots, repeat steps S1-S5 to obtain the deformation data caused only by braking force in different time slots with reduced dimensions. Based on the deformation data caused solely by braking force in different time slots with reduced dimensions, time series data of deformation caused solely by braking force are obtained.

8. A device for detecting the dynamic braking force of a drum, characterized in that, include: The acquisition module is used to acquire the real-time rotation speed of the roller when a vehicle is detected entering the roller area, and to activate the non-contact light field sensing system to simultaneously acquire the total three-dimensional deformation time series data and temperature change time series data of the roller area. The first deformation time series data determination module is used to input the temperature change time series data and the real-time rotation speed into the coupling system of the drum and wheel based on the finite element method, respectively, to obtain the deformation time series data caused by temperature change and the deformation time series data caused by the centrifugal force of the drum rotation. The second deformation timing data determination module is used to determine the deformation timing data caused only by braking force based on the total three-dimensional deformation timing data of the drum area, the deformation timing data caused by temperature changes, and the deformation timing data caused by the centrifugal force of drum rotation. The braking force estimation module is used to estimate the dynamic braking force based on the deformation time series data caused solely by braking force.

9. An electronic device, the device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor performs the steps of the method for detecting the dynamic braking force of the drum according to any one of claims 1-7.

10. A computer storage medium storing computer instructions thereon, characterized in that, When executed by the processor, this instruction implements the steps of the method for detecting the dynamic braking force of the drum as described in any one of claims 1-7.

Images