Vehicle chassis detection method based on multi-source data fusion optimization

Abstract

The invention discloses a vehicle chassis detection method based on multi-source data fusion optimization, and relates to the technical field of intelligent parking management. The method comprises the following steps: collecting a real-time turning plate angle, a driving motor operation current and an environment temperature; the driving current at the current angle is calculated in combination with the inherent parameters of the turning plate, and the over-current judgment threshold is dynamically adjusted after correction of a pre-trained environmental factor influence model; and comparing the real-time running current with the overcurrent threshold, and if so, judging that the chassis is contacted and controlling the motor to stop. Through angle-current-temperature multi-source data fusion, the problems that an existing method is insufficient in fixed over-current threshold value precision and is greatly influenced by temperature are solved, an external sensor is not needed, the method is adaptive to a complex environment and various turning plate structures, and the detection reliability and applicability are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a vehicle chassis detection method based on multi-source data fusion optimization. Background Technology

[0002] Intelligent parking space managers are typically installed in on-street public paid parking spaces. Their core function is to manage parking fees and ensure proper vehicle parking. When a vehicle's parking time in a designated space exceeds the preset free parking time, a flap on the manager automatically rises to lock the vehicle, preventing unauthorized departure and ensuring the orderly operation of on-street public parking fees. Because different vehicle models have significantly different chassis heights, to avoid damage to the chassis from excessive flap raising while ensuring the flap is high enough for effective locking, the intelligent parking space manager must accurately detect the vehicle's chassis height and stop raising the flap when it reaches the appropriate height to contact the vehicle's chassis. This is a crucial requirement for ensuring the safe and stable operation of the intelligent parking space manager and balancing locking effectiveness with vehicle safety.

[0003] Currently, the industry mainly employs two methods to address the aforementioned chassis detection and flap start-stop requirements. The first method involves adding a switch-type sensor to the top of the flap. This sensor directly detects whether the flap is in contact with the vehicle chassis, triggering a command to stop the flap from rising. However, this method has significant application drawbacks. Because the intelligent parking space manager is installed in outdoor public areas, the operating environment is extremely complex. It not only faces long-term exposure to wind, sun, rain, dust accumulation, and freezing temperatures, but is also susceptible to external interference such as impacts from debris kicked up by passing vehicles, scratches and bumps from pedestrians or non-motorized vehicles, and equipment misalignment caused by road subsidence. This leads to frequent malfunctions of the switch-type sensor, including decreased sensitivity, poor contact, short circuits, and open circuits. The equipment has high maintenance costs and poor stability, making it unreliable to meet the long-term requirements for chassis detection and flap start-stop, thus limiting its practical application. The second method determines whether the flap is in contact with the chassis by detecting changes in the operating current of the flap drive motor. This method does not require additional external sensors, has stronger adaptability, and is currently the mainstream practice in the industry. However, this method still has a technical pain point of insufficient accuracy. The core reason is that the flap of the intelligent parking manager adopts a circular motion mode. The driving force required during the flap's raising process is not a constant value, but a function that changes with the flap's raising angle. Different flap raising angles result in significant differences in the corresponding torque requirements, which in turn causes the operating current of the drive motor to fluctuate regularly with the flap's angle. If only a fixed overcurrent change threshold is set as the judgment criterion, it is impossible to accurately distinguish whether the change in motor current is due to the change in driving force caused by the change in flap angle or the change in current caused by the increased resistance after the flap contacts the vehicle chassis, which is prone to misjudgment.

[0004] Furthermore, changes in outdoor ambient temperature (including ground temperature) exacerbate the aforementioned accuracy issues, becoming an additional interference factor that existing current detection methods cannot avoid. Although the main structure of the flap is buried underground, its exposed drive motor and circuit components, as well as the underground flap shaft and sensors, are still affected by temperature changes: On the one hand, the resistance of the internal windings of the drive motor changes significantly with temperature fluctuations, increasing in low temperatures and decreasing in high temperatures. Even if the flap is not in contact with the chassis and there is no additional resistance, the current required for normal motor operation at the same angle will shift due to temperature changes, causing the judgment benchmark of the fixed overcurrent threshold to fail. On the other hand, temperature changes cause thermal expansion and contraction of the flap body, slightly altering the flap's inherent parameters such as length and center of gravity. At the same time, low temperatures may cause the lubricating oil at the shaft to solidify, and high temperatures may cause changes in component gaps. These will all change the driving force and motion resistance required for the flap to rise normally, further deviating the matching relationship between current and driving force and increasing the risk of misjudgment. In addition, extreme high and low temperatures will also affect the detection accuracy of the angle sensor (if configured), indirectly amplifying the error in current judgment.

[0005] In summary, the existing chassis detection and flap start / stop control methods of intelligent parking space managers have inherent defects such as easy failure of external sensors and inaccurate fixed overcurrent thresholds. They also face additional interference from changes in ambient temperature, making it difficult to meet the actual use requirements for detection accuracy. They cannot balance the effectiveness of locking the vehicle and the safety of the vehicle. There is an urgent need for a more accurate and reliable detection and control solution to solve the problems of the existing technology. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a vehicle chassis detection method based on multi-source data fusion optimization, the method comprising:

[0007] The real-time tilt angle of the berth manager and the real-time operating current of the drive motor are collected.

[0008] Obtain the drive current at the real-time flip angle;

[0009] Collect real-time ambient temperature, input the real-time ambient temperature and the inherent parameters of the flap into the pre-trained environmental factor influence model, and obtain the driving current under the corrected real-time flap angle;

[0010] The overcurrent judgment threshold is dynamically adjusted based on the drive current under the corrected real-time flip angle.

[0011] By comparing the real-time detected drive motor operating current with the dynamically adjusted overcurrent judgment threshold, when the operating current exceeds the overcurrent judgment threshold, it is determined that the flap has contacted the vehicle chassis, and the drive motor is controlled to stop running, so that the flap stops rising.

[0012] Furthermore, the process of obtaining the driving current under the real-time flip angle is as follows: the inherent parameters of the flip are obtained in advance, including the flip weight, flip length, and distance from the flip center of gravity to the rotating shaft;

[0013] Combining the real-time flip-up angle with inherent parameters, the driving force required for the flip-up to rise normally at the current angle is calculated using a mechanical formula: Driving force = Flip-up weight × Standard gravitational acceleration × Distance from the flip-up center of gravity to the pivot axis × Cosine of the real-time flip-up angle ÷ Flip-up length;

[0014] The driving current is calculated based on the mapping relationship formula between driving force and driving motor current. The mapping relationship formula is: driving current = driving force × motor torque coefficient ÷ (motor operating efficiency × motor rated voltage), where the motor torque coefficient is a preset constant, the motor operating efficiency is a preset constant, and the motor rated voltage is a preset constant.

[0015] Furthermore, the pre-training process of the environmental factors affecting the model specifically includes:

[0016] Multiple consecutive temperature ranges can be set, each covering a preset temperature range, with the temperature span of each range remaining consistent.

[0017] In each temperature range, the control flap is gradually raised from the initial angle to the maximum angle, and data is recorded at fixed angle intervals. The data includes the current flap angle, flap movement resistance, normal operating current of the drive motor, and angle sensor detection value.

[0018] The above data were continuously collected in different temperature ranges to construct a historical correlation database of temperature-angle-driving force-normal current.

[0019] The data in the historical correlation database is preprocessed, and the correction coefficients corresponding to each temperature range are determined based on the preprocessed data. The model is then constructed and trained to obtain the pre-trained environmental factor influence model.

[0020] Furthermore, the data in the historical correlation database is preprocessed, and the correction coefficients for each temperature range are determined based on the preprocessed data. The specific steps are as follows:

[0021] Outlier removal criteria are used to remove outliers from the historical relational database, and noise filtering algorithms are used to filter noisy data to obtain a preprocessed dataset.

[0022] Based on the preprocessed dataset, the correction coefficients for each temperature range are calculated and expressed as follows:

[0023] Motor resistance correction factor = Measured motor resistance in the target temperature range ÷ Rated motor resistance at standard temperature;

[0024] Flip-plate length correction factor = Measured value of flip-plate length in the target temperature range ÷ Design value of flip-plate length at standard temperature;

[0025] Correction factor for the center of gravity distance of the flapper = Measured value of the distance from the center of gravity of the flapper to the shaft under the target temperature range ÷ Design value of the distance from the center of gravity of the flapper to the shaft under the standard temperature;

[0026] Shaft resistance correction factor = Measured shaft resistance under the target temperature range ÷ Theoretical shaft resistance at the corresponding angle under standard temperature;

[0027] The correction factors corresponding to each temperature range are compiled into a correction factor table.

[0028] Furthermore, the specific steps for building and training the model are as follows: using ambient temperature and flap angle as model input variables, and the corrected driving current as model output variable, a backpropagation neural network model is constructed.

[0029] The model has two hidden layers, each with a preset number of neurons, and uses the linear rectified activation function.

[0030] The data in the historical association database is divided into training set and validation set according to a preset ratio; the output variable corresponding to each data point in the training set is calculated, and the output variable is the corrected driving current.

[0031] Corrected drive current = (Flipper weight × Standard gravitational acceleration × Distance from flipper center of gravity to shaft × Flipper center of gravity distance correction coefficient × Cosine of real-time flipper angle ÷ (Flipper length × Flipper length correction coefficient) × Motor torque coefficient) ÷ (Motor operating efficiency × Motor rated voltage × Motor resistance correction coefficient) + (Theoretical shaft resistance at the corresponding angle under standard temperature × Shaft resistance correction coefficient × Motor torque coefficient) ÷ (Motor operating efficiency × Motor rated voltage × Flipper length);

[0032] The input and output variables from the training set are input into the backpropagation neural network model, and the model weights are updated through the backpropagation algorithm.

[0033] Furthermore, the steps of building and training the model also include:

[0034] After each training round, the validation set data is input into the model, and the deviation between the corrected driving current output by the model and the measured normal operating current in the validation set is calculated. The deviation evaluation index is the mean absolute error, and the calculation formula is: mean absolute error = sum of the absolute values ​​of the differences between the model output value and the measured value of all data in the validation set ÷ total number of data in the validation set.

[0035] The model hyperparameters are dynamically adjusted based on the mean absolute error index, and the training is repeated iteratively until the mean absolute error of the model stabilizes within the preset deviation threshold for multiple consecutive rounds, at which point the model converges.

[0036] Furthermore, the overcurrent judgment threshold is dynamically adjusted based on the drive current under the corrected real-time flip angle. Specifically, an adaptive redundancy value is set, where the adaptive redundancy value = the corrected drive current × the redundancy coefficient, and the redundancy coefficient is a preset proportional range, which is dynamically adjusted according to the flip material strength, drive motor power and vehicle chassis protection threshold; the overcurrent judgment threshold = the drive current under the corrected real-time flip angle + the adaptive redundancy value.

[0037] Furthermore, the historical correlation database of temperature-angle-driving force-normal current supports dynamic updates, specifically including: real-time acquisition of the berth manager's operating data, including ambient temperature, tilting angle, driving current, tilting motion resistance, and measured values ​​of tilting inherent parameters; after preprocessing the newly acquired data, determining whether it belongs to an existing temperature range: if it does, replacing the old data under the same angle within that range; if it does not, adding a new temperature range and the corresponding correction coefficient; for every preset number of newly added valid data, iteratively training the environmental factor influence model based on the updated database until the model's mean absolute error stabilizes within a preset deviation threshold.

[0038] A vehicle chassis detection device based on multi-source data fusion optimization, used in the above method, includes:

[0039] Angle acquisition component: Installed at the hinge of the flip panel, used to acquire the flip panel angle in real time and output the angle detection signal;

[0040] Current acquisition component: connected in series in the power supply circuit of the drive motor, used to acquire the motor operating current in real time and output the current detection signal, with the measurement range covering the motor operating current range; Temperature acquisition component: including temperature sensing element, used to acquire real-time ambient temperature signal.

[0041] Data storage component: used to store the inherent parameters of the flap, the historical correlation database of temperature-angle-driving force-normal current, the correction coefficient table, and the pre-trained environmental factor influence model;

[0042] The calculation and control component is electrically connected to the angle acquisition component, current acquisition component, temperature acquisition component, data storage component, and drive motor, respectively, and is used to perform drive force calculation, drive current mapping, model correction, overcurrent threshold adjustment, and motor start-stop control logic.

[0043] Drive component: Receives start / stop commands from the computing control component, controls the forward and reverse rotation and stop operation of the drive motor, and realizes the lifting and lowering action of the flap.

[0044] The positive and progressive effects of this invention are as follows:

[0045] This invention fuses multi-source data (angle, current, temperature, etc.) and dynamically adjusts the overcurrent judgment threshold by combining the inherent parameters of the flap and the influence of environmental factors, accurately distinguishing the causes of current changes and effectively avoiding false positives and false negatives. It significantly enhances detection stability in complex outdoor temperature environments by calibrating the temperature-induced effects of motor resistance, flap parameters, and shaft resistance using a temperature correction coefficient. Eliminating the need for external switch-type sensors avoids the risk of failures caused by harsh outdoor environments, and further improves long-term operational reliability through dynamic updates of the historical database and iterative model training. Simultaneously, it is compatible with flaps of different structures such as curved and telescopic flaps, and its calculation model is optimized through dynamic angle calibration, exhibiting broad adaptability and comprehensively improving the detection accuracy, environmental adaptability, and practical application value of intelligent parking space managers. Attached Figure Description

[0046] Figure 1 This is a flowchart of the steps of the present invention. Detailed Implementation

[0047] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0048] This embodiment discloses a vehicle chassis detection method and device based on multi-source data fusion optimization, applicable to intelligent parking management scenarios in on-street public paid parking spaces. Its core function is to address the insufficient accuracy of flap contact detection caused by differences in chassis height among different vehicle models and fluctuations in ambient temperature. The technical solution of this invention will be described in detail below, combining specific equipment selection, parameter configuration, and operation procedures.

[0049] Reference Figure 1 A vehicle chassis detection method based on multi-source data fusion optimization is disclosed. The method includes: acquiring the real-time tilting angle of the parking space manager and the real-time operating current of the drive motor; obtaining the drive current under the real-time tilting angle; acquiring the real-time ambient temperature, inputting the real-time ambient temperature and the inherent parameters of the tilting into a pre-trained environmental factor influence model, and obtaining the corrected drive current under the real-time tilting angle; dynamically adjusting the overcurrent judgment threshold based on the corrected drive current under the real-time tilting angle; comparing the real-time detected drive motor operating current with the dynamically adjusted overcurrent judgment threshold, and when the operating current exceeds the overcurrent judgment threshold, determining that the tilting has contacted the vehicle chassis, controlling the drive motor to stop running, and stopping the tilting from rising.

[0050] Furthermore, the intelligent parking space manager is deployed in public paid parking spaces on urban roads. The main body of the flap is embedded in a pre-set groove in the parking space ground, with only the top of the flap flush with the ground to avoid affecting the normal entry and exit of vehicles. All data acquisition and computing control components are integrated inside the manager body and encapsulated in a waterproof, dustproof, and corrosion-resistant shell to adapt to complex outdoor environments.

[0051] Furthermore, the process of obtaining the drive current at the real-time flip angle is as follows: The inherent parameters of the flip are pre-obtained, including the flip weight, flip length, and distance from the flip center of gravity to the pivot. Combining the real-time flip angle with the inherent parameters, the driving force required for the flip to rise normally at the current angle is calculated using a mechanical formula: Driving force = Flip weight × Standard gravitational acceleration × Distance from flip center of gravity to pivot × Cosine of real-time flip angle ÷ Flip length. The drive current is then calculated based on the mapping relationship between the driving force and the drive motor current: Drive current = Driving force × Motor torque coefficient ÷ (Motor operating efficiency × Motor rated voltage), where the motor torque coefficient, motor operating efficiency, and motor rated voltage are preset constants.

[0052] In this embodiment, the specific parameters of the flap are: flap weight 15kg, flap length 80cm, distance from the flap center of gravity to the pivot 40cm; standard gravitational acceleration is 9.8m / s²; the drive motor is a DC geared motor with a preset torque coefficient of 0.8N・m / A, a preset operating efficiency of 0.9, and a preset rated voltage of 24V. For example, when the real-time flap angle is 30°, the cosine of the real-time flap angle is 0.866. Substituting into the mechanics formula, the driving force is calculated as 15×9.8×0.4×0.866÷0.8≈63.7N; then, the driving current is calculated using the mapping formula as 63.7×0.8÷(0.9×24)≈2.36A, that is, the driving current corresponding to the normal raising of the flap at the current angle is approximately 2.36A.

[0053] Furthermore, the pre-training process of the environmental factor influence model specifically includes: setting multiple continuous temperature intervals, each covering a preset temperature range, with a consistent temperature span within each interval; controlling the flap to gradually rise from its initial angle to its maximum angle within each temperature interval, recording data at fixed angle intervals, including the current flap angle, flap movement resistance, normal operating current of the drive motor, and angle sensor detection values; continuously collecting the above data under different temperature intervals to construct a historical correlation database of temperature-angle-driving force-normal current; preprocessing the data in the historical correlation database, determining the correction coefficients corresponding to each temperature interval based on the preprocessed data, constructing the model, and training it to obtain the pre-trained environmental factor influence model.

[0054] In this embodiment, the preset temperature range is -40℃ to 60℃, divided into 20 consecutive intervals in 5℃ increments (-40℃~-35℃, -35℃~-30℃…55℃~60℃). The initial angle of the flap is 0° (horizontally touching the ground), and the maximum angle is 90° (vertically locked). The fixed angle interval is 1°, meaning data is recorded every 1° increase. During data acquisition, the flap's movement resistance is indirectly measured by a torque sensor, and the normal operating current of the drive motor is recorded in real time by a series current acquisition component. The angle sensor's detection value is consistent with the output signal of the angle acquisition component. The acquisition duration covers typical temperature periods in different seasons, ensuring at least 50 sets of valid data for each temperature interval. The final historical database contains approximately 10,000 valid data records.

[0055] Furthermore, the data in the historical correlation database is preprocessed. The specific steps for determining the correction coefficients corresponding to each temperature range based on the preprocessed data are as follows: outlier removal criteria are used to remove outliers from the historical correlation database, and noise data is filtered through a noise filtering algorithm to obtain a preprocessed dataset; based on the preprocessed dataset, the correction coefficients corresponding to each temperature range are calculated, expressed as follows: motor resistance correction coefficient = measured motor resistance in the target temperature range ÷ rated motor resistance at the standard temperature; flap length correction coefficient = measured flap length in the target temperature range ÷ designed flap length at the standard temperature; flap center of gravity distance correction coefficient = measured distance from flap center of gravity to shaft in the target temperature range ÷ designed distance from flap center of gravity to shaft at the standard temperature; shaft resistance correction coefficient = measured shaft resistance in the target temperature range ÷ theoretical shaft resistance at the corresponding angle at the standard temperature; the above correction coefficients corresponding to each temperature range are compiled into a correction coefficient table.

[0056] In this embodiment, the outlier removal criterion adopts the 3σ criterion, which calculates the mean μ and standard deviation σ of each data dimension and removes data that exceeds the range of [μ-3σ, μ+3σ]. The noise filtering algorithm uses the moving average method with a window size of 5, replacing the original data with the mean of 5 adjacent data sets to reduce random noise interference. The standard temperature is set to 25℃, at which temperature the measured rated resistance of the motor is 5Ω, the designed length of the flap is 80cm, the designed distance from the center of gravity of the flap to the shaft is 40cm, and the theoretical resistance of the shaft at a corresponding 30° angle is 5N. Taking the target temperature range of -10℃ to -5℃ as an example, the measured motor resistance is 6.2Ω, the measured flap length is 79.9cm, the measured distance from the flap center of gravity to the shaft is 39.9cm, and the measured shaft resistance is 7.5N. The calculated correction coefficients are: motor resistance = 6.2 ÷ 5 = 1.24, flap length = 79.9 ÷ 80 = 0.99875, flap center of gravity distance = 39.9 ÷ 40 = 0.9975, and shaft resistance = 7.5 ÷ 5 = 1.5. The above coefficients and the calculation results for other temperature ranges are compiled into a structured correction coefficient table and stored in the data storage component.

[0057] Further, the specific steps for building and training the model are as follows: Using ambient temperature and flap angle as input variables, and the corrected drive current as output variable, a backpropagation neural network model is constructed. The model has two hidden layers, each with a preset number of neurons, and uses a linear rectified activation function. Data from the historical database is divided into training and validation sets according to a preset ratio. The output variable for each data point in the training set is calculated; the output variable is the corrected drive current. The corrected drive current = (flap weight × standard gravitational acceleration × distance from flap center of gravity to shaft × flap center of gravity distance correction coefficient × cosine of real-time flap angle ÷ (flap length × flap length correction coefficient) × motor torque coefficient) ÷ (motor operating efficiency × motor rated voltage × motor resistance correction coefficient) + (theoretical shaft resistance at the corresponding angle under standard temperature × shaft resistance correction coefficient × motor torque coefficient) ÷ (motor operating efficiency × motor rated voltage × flap length). The input and output variables from the training set are input into the backpropagation neural network model, and the model weights are updated using the backpropagation algorithm.

[0058] In this embodiment, the specific architecture of the backpropagation neural network model is as follows: the input layer has a dimension of 2 (corresponding to the two input variables, ambient temperature and flap angle), the first hidden layer has 16 neurons, the second hidden layer has 8 neurons, and the output layer has a dimension of 1 (corresponding to the corrected driving current). The activation function used is the rectified linear activation function (ReLU) to enhance the model's nonlinear fitting ability. The data in the historical association database is divided into a training set and a validation set in a preset ratio of 7:3, that is, 7000 data points are used for model training and 3000 data points are used for performance validation. Taking a target temperature range of -10℃ to -5℃ and a flap angle of 30° as an example, substituting the corresponding coefficients in the correction coefficient table, the corrected drive current is calculated as follows: (15×9.8×0.4×0.9975×0.866÷(0.8×0.99875)×0.8)÷(0.9×24×1.24)+(5×1.5×0.8)÷(0.9×24×0.8)≈(15×9.8×0.4×0.9975) ×0.866÷0.799×0.8)÷(26.784)+(6)÷(17.28)≈(41.02÷0.799×0.8)÷26.784+0.347≈(41.06×0.8)÷26.784+0.347≈32.85÷26.784+0.347≈1.226+0.347≈1.573A, which is the output variable corresponding to the current input variable. Substitute the input variables of all training set data and the calculated output variables into the model, use stochastic gradient descent as the backpropagation algorithm, set the initial learning rate to 0.01, and iteratively update the model weight parameters.

[0059] Furthermore, the steps of building and training the model also include: after each training round, inputting validation set data into the model, calculating the deviation between the corrected driving current output by the model and the measured normal operating current in the validation set, and using the mean absolute error as the deviation evaluation index. The calculation formula is: mean absolute error = sum of the absolute values ​​of the differences between the model output value and the measured value of all data in the validation set ÷ total number of data in the validation set; dynamically adjusting the model hyperparameters based on the mean absolute error index, and repeating the training iteration until the mean absolute error of the model stabilizes within the preset deviation threshold for multiple consecutive rounds, at which point the model converges.

[0060] In this embodiment, the preset deviation threshold is 0.05A, the preset training epochs are 500, and the continuous stable epochs are set to 5. After each training epoch, 3000 validation set data are input into the model, and the mean absolute error is calculated: if the mean absolute error > 0.05A, the learning rate is adjusted to 0.9 times the current value, and 50 more training iterations are added; if 0.03A ≤ mean absolute error ≤ 0.05A, the learning rate remains unchanged; if the mean absolute error < 0.03A, the current hyperparameters remain unchanged. For example, after the 100th training round, the mean absolute error is 0.07A, so the learning rate is adjusted from 0.01 to 0.009; after the 200th training round, the mean absolute error drops to 0.045A, so the learning rate is kept at 0.009 and training continues; after the 320th to 324th training rounds, the mean absolute errors are 0.048A, 0.046A, 0.045A, 0.044A, and 0.043A, respectively, remaining stable within the deviation threshold for five consecutive rounds, indicating model convergence and the pre-trained environmental factor influence model is obtained.

[0061] Furthermore, the overcurrent judgment threshold is dynamically adjusted based on the drive current under the corrected real-time flip angle. Specifically, an adaptive redundancy value is set, where the adaptive redundancy value = the corrected drive current × the redundancy coefficient, and the redundancy coefficient is a preset proportional range, which is dynamically adjusted according to the flip material strength, drive motor power and vehicle chassis protection threshold; the overcurrent judgment threshold = the drive current under the corrected real-time flip angle + the adaptive redundancy value.

[0062] In this embodiment, the preset range of the redundancy coefficient is 5%-20%: when the flap is made of high-strength aluminum alloy (tensile strength ≥300MPa) and the drive motor power is 500W, the redundancy coefficient is 15%; if the flap is made of ordinary steel (tensile strength ≥235MPa) and the drive motor power is 300W, the redundancy coefficient is 10%; if the vehicle chassis protection threshold is low (e.g., the thickness of the bottom protective plate of the battery pack of a new energy vehicle is <3mm), the redundancy coefficient is 5%. For example, if the drive current at the corrected real-time flap angle is 1.573A, corresponding to a redundancy coefficient of 15%, then the adaptive redundancy value = 1.573 × 15% ≈ 0.236A, and the overcurrent judgment threshold = 1.573 + 0.236 ≈ 1.809A.

[0063] Furthermore, the historical correlation database of temperature-angle-driving force-normal current supports dynamic updates, specifically including: real-time acquisition of the berth manager's operating data, including ambient temperature, tilting angle, driving current, tilting motion resistance, and measured values ​​of tilting inherent parameters; after preprocessing the newly acquired data, determining whether it belongs to an existing temperature range: if it does, replacing the old data under the same angle within that range; if it does not, adding a new temperature range and the corresponding correction coefficient; for every preset number of newly added valid data, iteratively training the environmental factor influence model based on the updated database until the model's mean absolute error stabilizes within a preset deviation threshold.

[0064] In this embodiment, the preset number of valid data points is 1000. After real-time collection of operational data, outlier removal is performed using the 3σ criterion and noise filtering using the moving average method. If the ambient temperature is -3℃, which falls within the existing temperature range of -5℃ to 0℃, and the flip angle is 30°, then the old data corresponding to the flip angle of 30° within this range is replaced. If the ambient temperature is 62℃, which exceeds the original temperature range, a new temperature range of 60℃ to 65℃ is added, and the corresponding correction coefficients for this range are recalculated and added to the correction coefficient table. For every 1000 newly added valid data points, they are divided into a new training set and a new validation set in a 7:3 ratio. After merging with the original dataset, the environmental factor influence model is iteratively trained. During the iteration process, the hyperparameters are kept consistent with the initial training until the model's mean absolute error stabilizes within 0.05A, thus completing model optimization.

[0065] A vehicle chassis detection device based on multi-source data fusion optimization, used to implement the above method, includes:

[0066] Angle acquisition component: Installed at the flip plate shaft, it adopts an incremental encoder, model E6B2-CWZ6C, with a measurement range of 0°-360° and an accuracy of ±0.1°, and is used to acquire the flip plate angle in real time and output the angle detection signal;

[0067] Current acquisition component: A Hall current sensor, model ACS712, is connected in series in the power supply circuit of the drive motor. The measurement range is 0-50A and the accuracy is ±0.01A. It is used to acquire the motor operating current in real time and output the current detection signal. The measurement range covers the motor operating current range.

[0068] Temperature acquisition component: includes an NTC thermistor sensor, model MF52A, with a measurement range of -40℃ to 60℃ and an accuracy of ±0.5℃, used to acquire real-time ambient temperature signals;

[0069] Data storage components: A combination of SD card (capacity ≥32GB) and flash memory chip (capacity ≥16MB) is used to store the inherent parameters of the flip panel, the historical correlation database of temperature-angle-driving force-normal current, the correction coefficient table, and the pre-trained environmental factor influence model;

[0070] Computational control components: An STM32F407 microcontroller is selected and electrically connected to the angle acquisition component, current acquisition component, temperature acquisition component, data storage component and drive motor respectively. The detection signals of each acquisition component are read through the I2C communication interface, and the drive force calculation, drive current mapping, model correction, overcurrent threshold adjustment and motor start and stop control logic are executed based on the built-in algorithm.

[0071] Drive components: including a relay module (model HF3FF / 012-1ZS) and a motor drive chip (model L298N), which receive start and stop commands from the computing control component, control the forward and reverse rotation and stop operation of the drive motor, and realize the lifting and lowering action of the flap. The drive voltage is matched with the rated voltage of the motor at 24V.

[0072] In practical applications, when a vehicle enters a paid public parking space and remains parked for more than the preset free time (e.g., 15 minutes), the calculation and control component triggers the flap lifting process: the angle acquisition component collects the flap angle in real time (initially 0°), the current acquisition component collects the initial operating current of the drive motor, and the temperature acquisition component collects the real-time ambient temperature (e.g., 28°C); the calculation and control component calls the flap's inherent parameters from the data storage component, calculates the drive current in combination with the real-time flap angle, inputs the ambient temperature and flap's inherent parameters into a pre-trained environmental factor influence model, and obtains the corrected drive current; after dynamically adjusting the overcurrent judgment threshold, the motor operating current is compared with the overcurrent judgment threshold in real time. When the flap is raised to contact the vehicle chassis, the motor operating current suddenly increases to exceed the overcurrent judgment threshold, the calculation and control component immediately outputs a stop command, the drive component controls the drive motor to stop running, the flap stops rising, and the vehicle locking action is completed.

[0073] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A vehicle chassis detection method based on multi-source data fusion optimization, characterized in that, The method includes: The real-time tilt angle of the berth manager and the real-time operating current of the drive motor are collected. Obtain the drive current at the real-time flip angle; Collect real-time ambient temperature, input the real-time ambient temperature and the inherent parameters of the flap into the pre-trained environmental factor influence model, and obtain the driving current under the corrected real-time flap angle; The overcurrent judgment threshold is dynamically adjusted based on the drive current under the corrected real-time flip angle. By comparing the real-time detected drive motor operating current with the dynamically adjusted overcurrent judgment threshold, when the operating current exceeds the overcurrent judgment threshold, it is determined that the flap has contacted the vehicle chassis, and the drive motor is controlled to stop running, so that the flap stops rising.

2. The vehicle chassis detection method based on multi-source data fusion optimization according to claim 1, characterized in that, The process of obtaining the driving current under the real-time flip angle is as follows: the inherent parameters of the flip are obtained in advance, including the flip weight, flip length, and distance from the flip center of gravity to the rotating shaft; Combining the real-time flip-up angle with inherent parameters, the driving force required for the flip-up to rise normally at the current angle is calculated using a mechanical formula: Driving force = Flip-up weight × Standard gravitational acceleration × Distance from the flip-up center of gravity to the pivot axis × Cosine of the real-time flip-up angle ÷ Flip-up length; The driving current is calculated based on the mapping relationship formula between driving force and driving motor current. The mapping relationship formula is: driving current = driving force × motor torque coefficient ÷ (motor operating efficiency × motor rated voltage), where the motor torque coefficient is a preset constant, the motor operating efficiency is a preset constant, and the motor rated voltage is a preset constant.

3. The vehicle chassis detection method based on multi-source data fusion optimization according to claim 1, characterized in that, The pre-training process of the environmental factors influence model specifically includes: Multiple consecutive temperature ranges can be set, each covering a preset temperature range, with the temperature span of each range remaining consistent. In each temperature range, the control flap is gradually raised from the initial angle to the maximum angle, and data is recorded at fixed angle intervals. The data includes the current flap angle, flap movement resistance, normal operating current of the drive motor, and angle sensor detection value. The above data were continuously collected in different temperature ranges to construct a historical correlation database of temperature-angle-driving force-normal current. The data in the historical correlation database is preprocessed, and the correction coefficients corresponding to each temperature range are determined based on the preprocessed data. The model is then constructed and trained to obtain the pre-trained environmental factor influence model.

4. The vehicle chassis detection method based on multi-source data fusion optimization according to claim 3, characterized in that, The specific steps for preprocessing data in the historical correlation database and determining the correction coefficients for each temperature range based on the preprocessed data are as follows: Outlier removal criteria are used to remove outliers from the historical relational database, and noise filtering algorithms are used to filter noisy data to obtain a preprocessed dataset. Based on the preprocessed dataset, the correction coefficients for each temperature range are calculated and expressed as follows: Motor resistance correction factor = Measured motor resistance in the target temperature range ÷ Rated motor resistance at standard temperature; Flip-plate length correction factor = Measured value of flip-plate length in the target temperature range ÷ Design value of flip-plate length at standard temperature; Correction factor for the center of gravity distance of the flapper = Measured value of the distance from the center of gravity of the flapper to the shaft under the target temperature range ÷ Design value of the distance from the center of gravity of the flapper to the shaft under the standard temperature; Shaft resistance correction factor = Measured shaft resistance under the target temperature range ÷ Theoretical shaft resistance at the corresponding angle under standard temperature; The correction factors corresponding to each temperature range are compiled into a correction factor table.

5. The vehicle chassis detection method based on multi-source data fusion optimization according to claim 4, characterized in that, The specific steps for building and training the model are as follows: using ambient temperature and flap angle as model input variables and the corrected driving current as model output variables, a backpropagation neural network model is constructed. The model has two hidden layers, each with a preset number of neurons, and uses the linear rectified activation function. The data in the historical relational database is divided into training set and validation set according to a preset ratio; Calculate the output variable corresponding to each data point in the training set. The output variable is the corrected driving current. Corrected drive current = (Flipper weight × Standard gravitational acceleration × Distance from flipper center of gravity to shaft × Flipper center of gravity distance correction coefficient × Cosine of real-time flipper angle ÷ (Flipper length × Flipper length correction coefficient) × Motor torque coefficient) ÷ (Motor operating efficiency × Motor rated voltage × Motor resistance correction coefficient) + (Theoretical shaft resistance at the corresponding angle under standard temperature × Shaft resistance correction coefficient × Motor torque coefficient) ÷ (Motor operating efficiency × Motor rated voltage × Flipper length); The input and output variables from the training set are input into the backpropagation neural network model, and the model weights are updated through the backpropagation algorithm.

6. The vehicle chassis detection method based on multi-source data fusion optimization according to claim 5, characterized in that, The steps of building and training a model also include: After each training round, the validation set data is input into the model, and the deviation between the corrected driving current output by the model and the measured normal operating current in the validation set is calculated. The deviation evaluation index is the mean absolute error, and the calculation formula is: mean absolute error = sum of the absolute values ​​of the differences between the model output value and the measured value of all data in the validation set ÷ total number of data in the validation set. The model hyperparameters are dynamically adjusted based on the mean absolute error index, and the training is repeated iteratively until the mean absolute error of the model stabilizes within the preset deviation threshold for multiple consecutive rounds, at which point the model converges.

7. The vehicle chassis detection method based on multi-source data fusion optimization according to claim 1, characterized in that, The overcurrent judgment threshold is dynamically adjusted based on the drive current under the corrected real-time flip angle. Specifically, an adaptive redundancy value is set, where the adaptive redundancy value = the corrected drive current × the redundancy coefficient, and the redundancy coefficient is a preset proportional range, which is dynamically adjusted according to the flip material strength, drive motor power and vehicle chassis protection threshold. Overcurrent detection threshold = drive current at the corrected real-time flip angle + adaptive redundancy value.

8. The vehicle chassis detection method based on multi-source data fusion optimization according to claim 3, characterized in that, The historical correlation database of temperature, angle, driving force, and normal current supports dynamic updates. Specifically, it includes: real-time acquisition of the berth manager's operating data, including ambient temperature, tilting plate angle, driving current, tilting plate movement resistance, and measured values ​​of tilting plate inherent parameters; after preprocessing the newly acquired data, determining whether it belongs to an existing temperature range: if it does, replacing the old data for the same angle within that range; if it does not, adding a new temperature range and corresponding correction coefficients; and iteratively training the environmental factor influence model based on the updated database for each preset number of newly added valid data until the model's mean absolute error stabilizes within a preset deviation threshold.

Images