Thermostat stroke self-adaptive control method

Abstract

The invention relates to the technical field of engine heat management, and discloses a thermostat stroke self-adaptive control method. According to the method, multi-source working condition signals such as engine coolant temperature time sequence data, a load change rate and cylinder vibration spectrum characteristics are collected, and working condition intention vectors representing target temperature adjustment requirements and thermal management constraint conditions are generated through working condition intention analysis; and a thermostat dynamic response model containing a valve element displacement field equation, a cooling medium flow resistance characteristic function and a thermal inertia delay parameter is constructed, a target stroke control instruction is calculated and executed in combination with the working condition intention vector, and meanwhile the actual displacement track of the valve element and the cooling medium flow change rate are monitored. And generating a response efficiency evaluation index based on the displacement deviation value and the flow change rate, correcting a thermal inertia delay parameter of the model according to the response efficiency evaluation index, updating a control instruction, and forming a closed-loop self-adaptive adjustment mechanism to adapt to the complex working condition of the engine and optimize the thermal management effect.

Description

Technical Field

[0001] This invention relates to the field of engine thermal management technology, specifically a thermostat stroke adaptive control method. Background Technology

[0002] During engine operation, the performance of the thermal management system directly affects the engine's efficiency and lifespan. The thermostat, as the core control element of the thermal management system, plays a decisive role in the effectiveness of coolant temperature control due to its stroke adjustment accuracy and response speed. Currently, most mainstream thermostat control methods on the market employ a fixed threshold triggering mechanism, which controls the opening and closing of the valve core based on whether the coolant temperature reaches preset upper or lower thresholds. This control method is simple in structure and low in cost, and was widely used in early engine technology. However, as engines develop towards higher power, lower fuel consumption, and lower emissions, its limitations have become increasingly apparent. Fixed threshold control cannot adapt to the complex and ever-changing operating conditions of an engine. During actual operation, engine conditions change in real time due to factors such as driving habits, road conditions, and load. For example, when a vehicle starts and accelerates, the engine load increases sharply, and the coolant temperature rises rapidly; during high-speed cruising, the engine load is relatively stable, and the temperature change is gradual; when idling, the load is low, and the temperature rises slowly. Under different operating conditions, the engine's coolant temperature requirements vary significantly. Fixed threshold control can only adjust according to a preset single temperature standard and cannot dynamically adjust the temperature control target according to changes in operating conditions. This can easily lead to excessively high coolant temperatures under high load conditions, affecting engine power output, or excessively low temperatures under low load conditions, increasing fuel consumption and pollutant emissions. Existing thermostat control methods lack a dynamic correction mechanism for their own dynamic response characteristics. During long-term use, the internal valve core seals of the thermostat wear down, the flow resistance characteristics of the coolant change, and the thermal inertia between the cylinder block and coolant fluctuates due to factors such as operating environment and maintenance conditions. These changes cause deviations between the actual dynamic response of the thermostat and the initial design model. Most existing control methods calculate control commands based on fixed dynamic response models, failing to consider changes in model parameters during use. This leads to a mismatch between control commands and the actual movement of the valve core, resulting in valve core opening lag or over-adjustment. For example, when the internal flow resistance of the thermostat increases, the control commands calculated according to the original model cannot push the valve core to the expected stroke, resulting in insufficient coolant flow and decreased temperature control accuracy. When the thermal inertia delay parameter decreases, the control commands calculated by the original model may cause the valve core to over-adjust, resulting in frequent temperature fluctuations and exacerbating thermal fatigue damage to engine components. Existing control methods for evaluating thermostat response performance rely on a single dimension, typically using only the deviation between coolant temperature and target temperature as the evaluation indicator. This neglects the impact of key parameters such as the actual displacement trajectory of the valve core and the rate of change in coolant flow rate on control effectiveness. This one-dimensional evaluation approach fails to comprehensively reflect the actual operating state of the thermostat, making it difficult to accurately identify problems in the control process. Consequently, it hinders targeted optimization of the control strategy, further limiting improvements in thermostat control accuracy and response speed. Consequently, it fails to meet the high precision and adaptability requirements of modern engine thermal management systems. Summary of the Invention

[0003] The purpose of this invention is to provide a thermostat stroke adaptive control method to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides a thermostat stroke adaptive control method, the method comprising: Collect multi-source operating condition signals of engine operating status, including coolant temperature time-series data, engine load change rate and cylinder block vibration spectrum characteristics. The multi-source operating condition signals are processed by operating condition intent parsing to generate an engine operating condition intent vector, which represents the target temperature regulation requirements and thermal management constraints. A dynamic response model of the thermostat is constructed, which includes the valve core displacement field equation, the cooling medium flow resistance characteristic function, and the thermal inertia delay parameter. Calculate the target stroke control command based on the engine operating condition intention vector and the thermostat dynamic response model; The target stroke control command is executed to drive the thermostat valve core to move, and the actual displacement trajectory of the valve core and the rate of change of the cooling medium flow rate are monitored in real time. Based on the deviation between the actual displacement trajectory of the valve core and the target stroke control command, combined with the cooling medium flow rate change rate, a thermostat response performance evaluation index is generated. Based on the thermostat response performance evaluation index, the thermal inertia delay parameter in the thermostat dynamic response model is dynamically corrected; The target stroke control command is updated based on the corrected thermal inertia delay parameter to form a closed-loop adaptive adjustment mechanism.

[0005] Preferably, the step of performing condition intent parsing processing on the multi-source operating condition signals to generate an engine operating condition intent vector includes: Extract the rise slope, steady-state fluctuation range, and overshoot probability characteristics of the coolant temperature time-series data; Analyze the mapping relationship between the engine load change rate and the target temperature setpoint to generate a load-temperature coupling coefficient; Identify the characteristic frequency band energy distribution related to heat load in the cylinder block vibration spectrum characteristics; By integrating the rising slope, steady-state fluctuation range, overshoot probability characteristics, load-temperature coupling coefficient, and characteristic frequency band energy distribution, the engine operating condition intention vector is generated through a multi-dimensional weighted intention fusion algorithm.

[0006] Preferably, the construction of the thermostat dynamic response model includes: The valve core displacement field equation is established based on the mechanical structural parameters of the thermostat. The valve core displacement field equation includes the spring preload variable and the hydraulic actuation response function. The flow resistance characteristic function of the cooling medium under different opening degrees is calibrated, and the flow resistance characteristic function of the cooling medium includes a viscosity-temperature correction factor. The thermal inertia delay parameter was obtained through a step response experiment. The thermal inertia delay parameter includes the thermal conduction time constant of the medium and the heat storage coefficient of the metal. By combining the valve core displacement field equation, the cooling medium flow resistance characteristic function, and the thermal inertia delay parameter, a state-space expression for the thermostat dynamic response model is constructed.

[0007] Preferably, the step of calculating the target stroke control command based on the engine operating condition intention vector and the thermostat dynamic response model includes: The engine operating condition intention vector is input into a pre-trained stroke prediction model, which is constructed based on the thermostat dynamic response model. The trip prediction model outputs the initial trip control quantity and its corresponding expected flow rate change curve. The expected flow rate change curve is subjected to three-dimensional flow field reconstruction processing to generate a cooling medium temperature fluctuation matrix; The initial stroke control value is corrected based on the axial temperature gradient, radial heat exchange efficiency, and eddy current intensity coefficient in the cooling medium temperature fluctuation matrix. The corrected initial stroke control value is converted into a pulse width modulation signal to generate the target stroke control command.

[0008] Preferably, the step of performing three-dimensional flow field reconstruction processing on the expected flow rate change curve to generate a cooling medium temperature fluctuation matrix includes: The expected flow rate change curve is divided into multiple flow rate subsequences according to a time window; For each traffic subsequence, perform the following processing: A three-dimensional temperature field topology for cooling pipes is constructed based on fluid dynamics equations. The three-dimensional temperature field topology includes spatial data of axial temperature distribution field, radial heat diffusion field and tangential vortex field. The three-dimensional temperature field topology is thermodynamically coupled with the flow subsequence to generate the temperature field reconstruction result for the current time window. Thermal accumulation effect analysis is performed on the temperature field reconstruction results of continuous time windows to extract the axial temperature gradient, radial heat exchange efficiency and eddy current intensity coefficient.

[0009] Preferably, the step of generating a thermostat response performance evaluation index based on the deviation between the actual displacement trajectory of the valve core and the target stroke control command, combined with the cooling medium flow rate change rate, includes: Calculate the following delay time and overshoot of the actual displacement trajectory of the valve core relative to the target stroke control command; Analyze the deviation between the actual response rate and the expected rate of the cooling medium flow rate change. By integrating the following delay time, overshoot, and deviation, a dynamic response error coefficient is generated. Calculate the response baseline correction factor based on the ambient temperature and engine oil viscosity parameters in the historical control data; The dynamic response error coefficient is multiplied by the response benchmark correction factor to generate the thermostat response performance evaluation index.

[0010] Preferably, the step of dynamically correcting the thermal inertia delay parameter in the thermostat dynamic response model based on the thermostat response performance evaluation index includes: Establish an inverse mapping relationship between the thermal inertia delay parameter and the thermostat response performance evaluation index; The compensation amount for the thermal inertia delay parameter is calculated using the gradient descent algorithm; A first compensation weight is applied to the medium thermal conduction time constant in the thermal inertia delay parameter; A second compensation weight is applied to the metal heat storage coefficient, and the second compensation weight is related to the specific heat capacity of the engine block material; The thermal inertia delay parameter is updated based on the weighted compensation result.

[0011] Preferably, updating the target stroke control command based on the corrected thermal inertia delay parameter includes: The updated thermal inertia delay parameters are fed back to the thermostat dynamic response model; Recalculate the hydraulic actuation response function in the valve core displacement field equation; The engine operating condition intent vector, the updated hydraulic actuation response function, and the cooling medium flow resistance characteristic function are integrated through a multi-dimensional control vector fusion algorithm. Generate updated target travel control commands that include position feedforward compensation and flow feedback correction.

[0012] Preferably, the method further includes: Monitor sudden environmental temperature events and changes in the composition of the cooling medium; When a sudden change in ambient temperature is detected, a thermal stress compensation mechanism is triggered to perform thermal shock correction on the cooling medium temperature fluctuation matrix. When a change in the composition of the cooling medium is detected, the viscosity-temperature correction factor in the flow resistance characteristic function of the cooling medium is adjusted. The corrected cooling medium temperature fluctuation matrix and flow resistance characteristic function are input into the stroke prediction model to recalculate the target stroke control command.

[0013] Preferably, the method further includes: Collect deviation data between the actual temperature adjustment curve and the expected temperature curve within the preset verification period; The travel prediction model is trained using residual backpropagation based on the deviation data. Update the weight parameters and convolution kernel coefficients of the feature extraction layer in the trip prediction model; The updated travel prediction model will be applied to subsequent control cycles.

[0014] Compared with the prior art, the beneficial effects of the present invention are: By collecting multi-source operating condition signals of the engine, this method breaks away from the reliance on a single temperature signal in traditional fixed threshold control, enabling comprehensive capture of dynamic changes in engine operating conditions. The coolant temperature time-series data included in the multi-source operating condition signals reflects real-time temperature trends, the engine load change rate reflects fluctuations in operating load, and the cylinder block vibration spectrum characteristics indirectly reflect the engine's working intensity and stability. The combination of these three provides rich and comprehensive foundational data for subsequent operating condition intent analysis, allowing the operating condition intent vector to more accurately represent target temperature regulation requirements and thermal management constraints. This avoids the operating condition misjudgment problems caused by incomplete information acquisition in traditional control methods, laying the foundation for precise temperature control. In terms of control model construction, the dynamic response model of the thermostat constructed by this method includes the valve core displacement field equation, the cooling medium flow resistance characteristic function, and the thermal inertia delay parameter. Compared with the traditional model that only considers the simple correspondence between temperature and valve core displacement, this model is closer to the actual working mechanism of the thermostat. The valve core displacement field equation can accurately describe the motion trajectory of the valve core under different control commands, the cooling medium flow resistance characteristic function can reflect the influence of flow resistance changes on coolant flow rate, and the thermal inertia delay parameter considers the heat transfer delay effect between the cylinder block and the coolant. The synergistic effect of these three factors enables the model to more accurately simulate the dynamic response process of the thermostat under different operating conditions, providing a reliable basis for the accurate calculation of the target stroke control command and effectively reducing the control deviation caused by the mismatch between the model and the actual operating conditions. The method introduces a closed-loop adaptive adjustment mechanism, enabling real-time correction of the thermostat's dynamic response model and dynamic updates of control commands. During control, by real-time monitoring of the actual valve core displacement trajectory and the rate of change in cooling medium flow, and combining the deviation between the actual valve core displacement and the target stroke, a thermostat response performance evaluation index is generated. This index comprehensively evaluates the thermostat's operating status from two dimensions: displacement accuracy and flow adaptability. Compared to the traditional single temperature deviation evaluation method, it more accurately identifies problems in the control process. Based on this evaluation index, the thermal inertia delay parameter is dynamically corrected, ensuring that the model parameters remain consistent with the actual operating state of the thermostat, avoiding model parameter drift caused by long-term use. Furthermore, by updating the target stroke control command, it ensures that the valve core movement continuously meets the temperature regulation requirements under the current operating conditions, effectively improving the stability and adaptability of the control.

[0015] This method, through precise analysis of operating conditions and dynamic model correction, ensures that the coolant temperature is consistently maintained within the optimal range. Under high load conditions, it prevents engine power loss and component damage caused by overheating, while under low load conditions, it prevents increased fuel consumption and excessive pollutant emissions caused by underheating, thus balancing engine power performance, economy, and environmental friendliness. Furthermore, by reducing excessive valve core adjustment and frequent fluctuations, it lowers the wear rate of valve core seals, extends the thermostat's lifespan, reduces maintenance costs and replacement frequency, and also reduces the risk of thermal fatigue damage to engine components caused by frequent temperature fluctuations, further improving the overall reliability and durability of the engine. Attached Figure Description

[0016] Figure 1 This is a schematic diagram illustrating the working principle of the thermostat stroke adaptive control method described in this invention. Figure 2 A flowchart for generating engine operating condition intent vectors; Figure 3 A flowchart generated for the target stroke control command; Figure 4 A flowchart for generating thermostat response performance evaluation indicators; Figure 5 This is a flowchart for correcting thermal inertia delay parameters. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0018] Please see Figure 1 The present invention provides a thermostat stroke adaptive control method, the method comprising: acquisition of multi-source operating condition signals, analysis of operating condition intent, construction of a thermostat dynamic response model, calculation and execution of target stroke control commands, and closed-loop adaptive adjustment.

[0019] Multi-source operating condition signals include coolant temperature time-series data, engine load change rate, and cylinder block vibration spectrum characteristics. These signals are acquired in real time through a sensor system. Coolant temperature data comes from a temperature sensor, engine load change rate comes from the engine control unit output, and cylinder block vibration spectrum characteristics are obtained through an accelerometer. Operating condition intent parsing and processing analyzes the multi-source operating condition signals to generate an engine operating condition intent vector, which represents the target temperature regulation requirement and thermal management constraints. The thermostat dynamic response model includes the valve core displacement field equation, coolant flow resistance characteristic function, and thermal inertia delay parameter. The model is constructed based on physical principles and experimental data. The target stroke control command is calculated based on the engine operating condition intent vector and the thermostat dynamic response model, and drives the thermostat valve core movement through an actuator. The actual valve core displacement trajectory and coolant flow rate change are monitored in real time using displacement and flow sensors. Based on the deviation between the actual valve core displacement trajectory and the target stroke control command, combined with the coolant flow rate change, a thermostat response performance evaluation index is generated. This index is used to dynamically correct the thermal inertia delay parameter in the thermostat dynamic response model. The corrected thermal inertia delay parameter is used to update the target stroke control command, forming a closed-loop adaptive adjustment mechanism to ensure control accuracy and adaptability.

[0020] Example 1: See Figure 2 During the engine cold start phase, the coolant temperature time-series data shows a slow upward trend. PT100 temperature sensors installed in the engine block and cooling circuit collect temperature data at a sampling frequency of 10Hz. The data preprocessing module calculates the temperature difference between adjacent sampling points, generating a temperature change rate sequence. The average slope of consecutive positive difference segments in the sequence is taken as the rising slope characteristic. The steady-state fluctuation range is obtained by calculating the difference between the maximum and minimum values ​​within a 30-second window after the temperature data enters a plateau period. The overshoot probability characteristic is obtained by statistically analyzing the proportion of times an oscillation peak occurs after the temperature first exceeds the set threshold of 85℃ in historical data out of the total number of starts.

[0021] Engine load change rate data is acquired from the engine control unit via the CAN bus, with a sampling frequency of 20Hz. The mapping relationship between the load change rate and the target temperature setpoint is implemented using a two-dimensional lookup table. This table takes the load change rate as input and outputs the corresponding temperature compensation value. For example, when the load change rate exceeds 5% per second, the lookup table outputs a target temperature compensation value that increases by 2°C. The load-temperature coupling coefficient is generated from this compensation value after normalization, and its value ranges from 0 to 1.

[0022] Cylinder block vibration data were acquired by a triaxial accelerometer mounted on the side of the engine block, with a sampling frequency of 2 kHz. The raw vibration signal was bandpass filtered and then subjected to a Fast Fourier Transform (FFT) to obtain a spectrum. The heat load-related characteristic frequency band was determined to be in the range of 100-500 Hz, and the energy integral value in this band was calculated through numerical integration. The energy distribution characteristic is expressed as the percentage of energy in this frequency band relative to the total energy.

[0023] The multi-dimensional weighted intent fusion algorithm employs a linear weighting method, with the following weights: ramp rate feature weight 0.3, steady-state fluctuation range weight 0.2, overshoot probability feature weight 0.1, load-temperature coupling coefficient weight 0.25, and characteristic frequency band energy distribution weight 0.15. Each feature value is standardized and multiplied by its corresponding weight; the weighted sum constitutes the first dimension of the engine operating condition intent vector. The vector also includes the original values ​​of each feature as auxiliary dimensions, ultimately forming a six-dimensional vector. This vector is updated every 200 milliseconds and input to the downstream control module.

[0024] The construction of the thermostat dynamic response model begins with the acquisition of mechanical parameters. The thermostat spring stiffness coefficient, measured using a material testing instrument, is 8 N / mm, and the initial preload is 15 N. The measured diameter of the hydraulic actuation cylinder is 20 mm, and the piston stroke range is 0-10 mm. The valve core displacement field equation is constructed based on Newton's second law, considering the balance of spring force, hydraulic pressure, and friction. The spring preload variable changes linearly with the valve core displacement. The hydraulic actuation response function includes a pressure-flow characteristic equation, and a second-order differential equation is used to describe the valve core motion process.

[0025] The flow resistance characteristics of the cooling medium were calibrated on a dedicated test bench. A thermostat was installed in the test loop, and flow-pressure differential data were recorded every 10% of the opening, from 0% to 100%. The test medium was a 50% ethylene glycol aqueous solution with a temperature range from -40℃ to 120℃. The viscosity-temperature correction factor was obtained by fitting a viscosity-temperature curve, using an exponential decay function. The flow resistance characteristic function was ultimately expressed as a three-dimensional interpolation table of opening, pressure differential, and temperature.

[0026] The thermal inertia delay parameter was determined through a step response experiment. With the thermostat fully closed, a sudden application of maximum hydraulic pressure was performed, and the time it took for the valve core displacement to reach 63.2% of its full stroke was recorded as the mechanical response delay. The medium heat conduction time constant was obtained by measuring the coolant temperature response curve. A temperature step signal was applied at the radiator inlet, and the time required for the temperature at each measuring point to reach 63.2% of its stable value was recorded. The metal heat storage coefficient was calculated by measuring the heat absorbed by the cylinder block material under a unit temperature change.

[0027] The state-space representation is constructed as a four-dimensional system, with state variables including valve spool displacement, valve spool velocity, coolant temperature, and metal temperature. The system matrix includes mass, damping, and stiffness parameters; the input matrix includes hydraulic pressure and inlet temperature; and the output matrix consists of valve spool displacement and outlet temperature. The model is discretized using the forward Euler method with a step size of 10 milliseconds, consistent with the sampling period of the control system. The model parameters are updated online every 1 second, continuously refining the prediction accuracy using real-time data.

[0028] Example 2: See Figure 3 During engine operation, the engine operating condition intent vector is continuously updated every 200 milliseconds. This vector contains six dimensions of data, representing temperature change trends, stability requirements, and thermal load status. This data is transmitted to the prediction module of the electronic control unit via the controller area network bus. The pre-trained stroke prediction model employs a long short-term memory neural network architecture, containing three hidden layers, each with 128 neurons, and using ReLU as the activation function. The network input layer receives the six-dimensional operating condition vector, and the output layer generates two prediction values: the initial stroke control quantity and its corresponding time-flow curve. The initial stroke control quantity represents the target opening percentage of the thermostat valve core, ranging from 0% to 100%. The expected flow rate change curve is output as an array, containing the predicted flow rate values ​​at 20 sampling points per second for the next 5 seconds, with the flow rate unit being liters per minute.

[0029] The expected flow rate change curve is fed into the flow field reconstruction module. First, the 5-second curve is divided into 5 consecutive time windows, each 1 second long and containing 20 flow rate data points. The flow rate data within each time window is timestamped and synchronized with the engine speed signal. For the flow rate subsequence of the first time window, the computational fluid dynamics simulation begins initialization. The simulation model is based on a 3D geometric model of the engine cooling system, which includes components such as the cylinder block water jacket, water pump, thermostat cavity, and radiator piping. Tetrahedral elements are used for mesh generation, with a minimum element size of 0.5 mm, resulting in approximately 3.5 million mesh elements.

[0030] In the boundary condition settings, the inlet pressure is determined based on the pump characteristic curve, and the outlet pressure is set to atmospheric pressure. The k-epsilon model is selected for the turbulence model, and the wall function uses standard wall treatment. The temperature field is initialized using the current actual coolant temperature as the initial value for the entire field. The transient calculation uses a pressure-velocity coupled algorithm with a time step of 0.05 seconds, performing 20 iterations within each time step. During the calculation, the flow rate subsequence data is input into the simulation model as the inlet flow rate boundary condition.

[0031] The simulation outputs three-dimensional temperature field data for each time step. This data is stored as a structured array containing the coordinates and temperature values ​​of each grid node. Three characteristic parameters are extracted from this data: the axial temperature gradient is obtained by calculating the average temperature difference between adjacent nodes along the flow direction; the radial heat exchange efficiency is obtained by calculating the temperature uniformity index on the cross section perpendicular to the flow direction; and the eddy current intensity coefficient is output by the eddy current calculation module in the post-processing software, with the cross-sectional average value taken as the representative value.

[0032] These characteristic parameters are organized into a temperature fluctuation matrix, a 5×3 two-dimensional array. Rows represent time window sequences, and columns correspond to the axial temperature gradient, radial heat exchange efficiency, and eddy current intensity coefficient, respectively. The matrix data, along with the initial stroke control value, is fed into a correction algorithm. The correction algorithm employs a fuzzy logic system, taking the three characteristic parameters from the matrix as input and outputting adjustment coefficients for the stroke control value. Multiplying these adjustment coefficients by the initial stroke control value yields the corrected control value. For example, when the axial temperature gradient is large, the adjustment coefficients will be appropriately increased to enhance the cooling effect.

[0033] The corrected stroke control value is sent to the signal conversion module. This module converts the percentage opening value into a pulse width modulation (PWM) signal. The conversion relationship is achieved through a lookup table, corresponding to the pulse width required for different opening values. The generated PWM signal has a frequency of 100Hz, and the duty cycle varies linearly according to the control value: 0% opening corresponds to a 5% duty cycle, and 100% opening corresponds to a 95% duty cycle. This signal is amplified by the drive circuit and then output to the electromagnetic actuator of the thermostat.

[0034] The entire process is repeated in each control cycle. Real-time monitored coolant flow data is fed back to the system via a flow sensor installed in the thermostat outlet pipe, with a sampling frequency of 100Hz. The flow data is used to verify the accuracy of the expected flow rate change curve and serves as the basis for model updates in subsequent control cycles. Simultaneously, the results of the 3D flow field reconstruction are compared with the actual temperature field data measured by an infrared thermal imager to correct the boundary conditions and material parameters of the computational fluid dynamics model. This closed-loop verification mechanism ensures the reliability of the flow field prediction, enabling the temperature fluctuation matrix to accurately reflect the actual thermal state of the cooling system.

[0035] Example 3: See Figure 4 In the adaptive control of the thermostat stroke, the evaluation of the system response performance is a continuous and dynamic process. A displacement sensor monitors the actual movement trajectory of the thermostat valve core in real time at a sampling frequency of 1000Hz. This sensor employs the magnetostrictive principle and achieves a measurement accuracy of ±0.1 mm. Simultaneously, the control system generates target stroke control commands every 10 milliseconds, and these commands are stored in a circular buffer as digital signals.

[0036] In the comparison and analysis of the actual displacement trajectory and the target command, time alignment is performed first. Due to the fixed communication delay between sensor acquisition and control command output, a cross-correlation algorithm is used to synchronize the two data sequences. The synchronized data is divided into several analysis windows, each 200 milliseconds long and containing 20 sampling points. Within each analysis window, the following delay time of the valve core's actual displacement trajectory relative to the target stroke control command is calculated. This value is determined by finding the time offset of the maximum cross-correlation point between the actual displacement curve and the target command curve during the rising edge phase. The overshoot is calculated by taking the percentage relative deviation between the maximum actual displacement value and the target command value during the stable phase after each command step change.

[0037] Monitoring of the cooling medium flow rate change rate is achieved using a turbine flow meter installed in the cooling line downstream of the thermostat. The meter has a measurement range of 0-200 liters / minute and an accuracy class of 1.0. Flow data is sampled at a frequency of 500 Hz, and after digital filtering, the flow rate change rate per second is calculated. The actual response rate is obtained by calculating the average flow rate change rate within each analysis window, while the expected rate is derived from the theoretical value output by the control model. The deviation index is obtained by calculating the absolute value of the relative error between the actual and expected rates, and smoothed using a moving average method.

[0038] These parameters are fused to generate dynamic response error coefficients. The fusion process uses a weighted geometric mean method, mathematically expressed as:

[0039] in: It represents the dynamic response error coefficient, which is a dimensionless evaluation index; This represents the follow delay time, in seconds; Indicates the overshoot amount, entered as a percentage; It is the deviation of the rate of change of flow, which is a dimensionless ratio; , , These are the weighting coefficients for each parameter, ranging from 0 to 1, and satisfying the following conditions: The specific values ​​of the weighting coefficients are determined through historical data analysis, with the delay time weight typically set at 0.4, the overshoot weight at 0.3, and the deviation weight at 0.3.

[0040] The response baseline correction factor is calculated based on historical operating data. Ambient temperature data is obtained from a temperature sensor installed near the engine intake manifold, with a sampling interval of 1 second. Engine oil viscosity parameters are obtained through a temperature-based estimation model that considers the relationship between oil brand, usage duration, and temperature variation. Within a 24-hour historical data window, the average and fluctuation range of ambient temperature, as well as the trend of engine oil viscosity variation, are extracted. These parameters are input into a multiple regression model, outputting a correction factor ranging from 0.8 to 1.2. This factor reflects the degree to which environmental conditions affect the baseline performance of the thermostat.

[0041] Finally, the thermostat response performance evaluation index is obtained by multiplying the dynamic response error coefficient by the response baseline correction factor: This indicator This is a comprehensive, dimensionless numerical value. The closer the value is to 1, the closer the system response performance is to the ideal state; the greater the deviation from 1, the more the system performance needs adjustment. This evaluation metric is updated every 200 milliseconds and transmitted in real time to the model correction module as a basis for adjusting the thermostat dynamic response model parameters. The entire evaluation process adopts a pipelined processing architecture, with each calculation step performed in parallel to ensure the real-time nature and accuracy of the evaluation results. Simultaneously, all intermediate parameters and final metrics are stored in a historical database for long-term performance trend analysis and model optimization reference.

[0042] Example 4: During the thermostat travel adaptive control process, the system continuously monitors and calculates the thermostat response performance evaluation index. This index is updated every 200 milliseconds, and its value fluctuates between 0.5 and 1.5. When the index deviates from the baseline value of 1.0, a dynamic correction process for the thermal inertia delay parameter is triggered. At the start of the correction process, the system establishes an inverse mapping relationship between the thermal inertia delay parameter and the response performance evaluation index. This mapping is achieved through a two-dimensional lookup table, which takes the response performance evaluation index as input and outputs the corresponding parameter adjustment direction and magnitude. For example, when the evaluation index value is 1.2, the lookup table indicates that the medium heat conduction time constant needs to be reduced, with an adjustment magnitude of five percent of the current value; when the evaluation index value is 0.8, it indicates that the metal heat storage coefficient needs to be increased, with an adjustment magnitude of three percent.

[0043] Gradient descent is used to calculate precise compensation amounts. The algorithm uses the deviation of the response performance evaluation metric from the ideal value of 1.0 as the loss function, and iteratively searches for the parameter adjustments that minimize the loss function. In each iteration, the algorithm calculates the partial derivatives of the loss function with respect to each parameter, and determines the direction and step size of the parameter adjustments based on the direction of the derivatives. The iterative process continues until the loss function value falls below a set threshold or the maximum number of iterations is reached.

[0044] The allocation of compensation weights is based on the sensitivity analysis of the parameters' impact on system performance. The medium's heat conduction time constant is assigned the first compensation weight, with a value of 0.6, reflecting the dominant influence of the cooling medium's characteristics on the system's thermal inertia. The metal heat storage coefficient receives the second compensation weight, with a value of 0.4. This weight value is related to the specific heat capacity data of the engine block material; the higher the specific heat capacity, the higher the weight value. The actual compensation process is achieved through weighted calculation, multiplying the original compensation amount obtained by the gradient descent algorithm by the corresponding weight to obtain the weighted compensation value. These weighted compensation values ​​are applied to the current thermal inertia delay parameters to generate updated parameter values. The entire weighted compensation calculation process is completed within each control cycle, ensuring the real-time nature of parameter updates.

[0045] The updated thermal inertia delay parameters are immediately fed back into the thermostat's dynamic response model. Upon receiving the new parameters, the model first recalculates the hydraulic actuation response function in the valve core displacement field equation. This recalculation involves solving a set of differential equations, taking into account the impact of the updated thermal inertia parameters on the dynamic characteristics of the hydraulic system. The calculation results show that both the time constant and damping coefficient of the hydraulic actuation response function have changed accordingly.

[0046] The multi-dimensional control vector fusion algorithm then commences operation, receiving three main inputs: a real-time updated engine operating condition intent vector, a corrected hydraulic actuation response function output, and current cooling medium flow resistance characteristic function data. The algorithm employs a weighted fusion strategy, assigning different weight coefficients to each input, which are dynamically adjusted based on the current operating state. The fusion process produces a comprehensive control vector that includes position control feedforward and flow regulation feedback.

[0047] The final generated target stroke control command consists of two components: a position feedforward compensation term and a flow feedback correction term. The position feedforward compensation, based on model predictions, pre-calculates the required valve core position adjustment; the flow feedback correction, on the other hand, fine-tunes the control command based on real-time monitored cooling medium flow data. These two parts are combined linearly to form a complete control command.

[0048] The entire parameter correction and control command update process is completed within one control cycle, ensuring the system's rapid response capability. The updated control command is output to the thermostat actuator via pulse width modulation, driving the valve core movement. Simultaneously, the system continues to monitor the actual response effect. The weight allocation during the thermal inertia delay parameter correction process is shown in Table 1.

[0049] Table 1: Weighting of Thermal Inertia Parameter Compensation.

[0050]

[0051] The data in this table is dynamically adjusted according to engine operating conditions during actual operation. The weighting coefficients reflect the importance of each parameter in the overall thermal inertia characteristics, the adjustment sensitivity index shows the degree of impact of parameter changes on system performance, and the maximum adjustment range limits the range of parameter changes within a single control cycle. These values ​​are derived from a large amount of experimental data and theoretical analysis, ensuring that the system can quickly adapt to changing operating conditions while maintaining stability.

[0052] Example 5: During engine operation, ambient temperature is continuously monitored by a digital temperature sensor installed near the intake manifold. This sensor uses a PT1000 platinum resistance element, with a measurement range covering -40℃ to 150℃ and a sampling frequency of 10Hz. The monitoring system calculates the rate of temperature change in real time. When a rate of temperature change exceeding 0.5℃ per second is detected, it is determined to be an ambient temperature abrupt change event. At this time, the system automatically triggers a thermal stress compensation mechanism, which first calls the thermal shock model calculation module after activation.

[0053] The thermal shock model is established based on the thermal expansion coefficient and thermal conductivity characteristics of materials, considering the different material properties of the engine block, piping, and thermostat body. Model inputs include the current ambient temperature, abrupt change magnitude, abrupt change duration, and material parameters of each component. The calculation process uses the finite difference method to solve the transient heat conduction equation, predicting the thermal stress distribution generated within the metal components by sudden temperature changes. Based on the calculation results, the system generates a temperature compensation matrix, which contains correction coefficients for each dimension of the cooling medium temperature fluctuation matrix. These correction coefficients are applied to the real-time updated temperature fluctuation matrix, adjusting its axial temperature gradient, radial heat exchange efficiency, and eddy current intensity coefficient.

[0054] Cooling medium composition monitoring is achieved through an online viscosity sensor installed in the cooling circuit. This sensor uses a vibrating plate principle to measure the dynamic viscosity of the coolant, with a measurement accuracy of ±1%. When the detected viscosity value deviates from the reference value by more than 5%, the system determines that the cooling medium composition has changed. At this time, the viscosity-temperature correction factor adjustment program is automatically initiated. The adjustment process first queries a preset medium characteristic database, which stores viscosity-temperature characteristic curves of coolants with different ratios. Based on the currently measured viscosity value, the system selects the closest characteristic curve and extracts the corresponding correction parameters. The new viscosity-temperature correction factor is updated in the cooling medium flow resistance characteristic function, replacing the original correction factor.

[0055] The corrected cooling medium temperature fluctuation matrix and flow resistance characteristic function are fed into the stroke prediction model in real time. Upon receiving these updated parameters, the model immediately recalculates. The calculation process first normalizes the input data, then performs forward propagation calculations using a neural network model. The new calculation results include the updated initial stroke control quantity and the expected flow rate change curve. These data undergo further processing to ultimately generate a new target stroke control command.

[0056] The system is configured with a fixed verification cycle, typically 24 hours of engine operation or 1000 cumulative work cycles. At the end of each verification cycle, the system automatically initiates a performance evaluation program. This program collects all actual temperature regulation data and expected temperature data output by the control model within that cycle. The actual temperature data comes from temperature sensors located at the engine coolant outlet and thermostat outlet, with a sampling interval of 100 milliseconds. After data alignment, the temperature deviation value at each sampling point is calculated, and the average deviation, maximum deviation, and standard deviation of the deviation are statistically analyzed for the entire cycle.

[0057] These bias data are fed into the model training module, which uses the backpropagation algorithm to optimize the parameters of the trip prediction model. The training process aims to minimize the bias, adjusting the connection weights and bias parameters of the neural network model. The training dataset consists of historical data from the last 10 validation periods, updated using a sliding window approach. Each training iteration includes two stages: forward computation and error backpropagation, progressively optimizing the model parameters using gradient descent.

[0058] After training, the updated travel prediction model is immediately put into practical control. The new model has improved feature extraction capabilities and prediction accuracy, and can better adapt to the actual operating conditions of the engine. The model update process is fully automated, requiring no manual intervention, ensuring that the control system continuously maintains optimal performance. The entire system forms a complete adaptive control closed loop, continuously optimizing the thermostat's control effect through continuous monitoring, evaluation, and adjustment.

[0059] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0060] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A thermostat stroke adaptive control method, characterized in that, The method includes: Collect multi-source operating condition signals of engine operating status, including coolant temperature time-series data, engine load change rate and cylinder block vibration spectrum characteristics. The multi-source operating condition signals are processed by operating condition intent parsing to generate an engine operating condition intent vector, which represents the target temperature regulation requirements and thermal management constraints. A dynamic response model of the thermostat is constructed, which includes the valve core displacement field equation, the cooling medium flow resistance characteristic function, and the thermal inertia delay parameter. Calculate the target stroke control command based on the engine operating condition intention vector and the thermostat dynamic response model; The target stroke control command is executed to drive the thermostat valve core to move, and the actual displacement trajectory of the valve core and the rate of change of the cooling medium flow rate are monitored in real time. Based on the deviation between the actual displacement trajectory of the valve core and the target stroke control command, combined with the cooling medium flow rate change rate, a thermostat response performance evaluation index is generated. Based on the thermostat response performance evaluation index, the thermal inertia delay parameter in the thermostat dynamic response model is dynamically corrected; The target stroke control command is updated based on the corrected thermal inertia delay parameter to form a closed-loop adaptive adjustment mechanism.

2. The method according to claim 1, characterized in that, The step of performing condition intent parsing processing on the multi-source operating condition signals to generate an engine operating condition intent vector includes: Extract the rise slope, steady-state fluctuation range, and overshoot probability characteristics of the coolant temperature time-series data; Analyze the mapping relationship between the engine load change rate and the target temperature setpoint to generate a load-temperature coupling coefficient; Identify the characteristic frequency band energy distribution related to heat load in the cylinder block vibration spectrum characteristics; By integrating the rising slope, steady-state fluctuation range, overshoot probability characteristics, load-temperature coupling coefficient, and characteristic frequency band energy distribution, the engine operating condition intention vector is generated through a multi-dimensional weighted intention fusion algorithm.

3. The method according to claim 2, characterized in that, The construction of the thermostat dynamic response model includes: The valve core displacement field equation is established based on the mechanical structural parameters of the thermostat. The valve core displacement field equation includes the spring preload variable and the hydraulic actuation response function. The flow resistance characteristic function of the cooling medium under different opening degrees is calibrated, and the flow resistance characteristic function of the cooling medium includes a viscosity-temperature correction factor. The thermal inertia delay parameter was obtained through a step response experiment. The thermal inertia delay parameter includes the thermal conduction time constant of the medium and the heat storage coefficient of the metal. By combining the valve core displacement field equation, the cooling medium flow resistance characteristic function, and the thermal inertia delay parameter, a state-space expression for the thermostat dynamic response model is constructed.

4. The method according to claim 3, characterized in that, The step of calculating the target stroke control command based on the engine operating condition intention vector and the thermostat dynamic response model includes: The engine operating condition intention vector is input into a pre-trained stroke prediction model, which is constructed based on the thermostat dynamic response model. The trip prediction model outputs the initial trip control quantity and its corresponding expected flow rate change curve. The expected flow rate change curve is subjected to three-dimensional flow field reconstruction processing to generate a cooling medium temperature fluctuation matrix; The initial stroke control value is corrected based on the axial temperature gradient, radial heat exchange efficiency, and eddy current intensity coefficient in the cooling medium temperature fluctuation matrix. The corrected initial stroke control value is converted into a pulse width modulation signal to generate the target stroke control command.

5. The method according to claim 4, characterized in that, The step of performing three-dimensional flow field reconstruction processing on the expected flow rate change curve to generate a cooling medium temperature fluctuation matrix includes: The expected flow rate change curve is divided into multiple flow rate subsequences according to a time window; For each traffic subsequence, perform the following processing: A three-dimensional temperature field topology for cooling pipes is constructed based on fluid dynamics equations. The three-dimensional temperature field topology includes spatial data of axial temperature distribution field, radial heat diffusion field and tangential vortex field. The three-dimensional temperature field topology is thermodynamically coupled with the flow subsequence to generate the temperature field reconstruction result for the current time window. Thermal accumulation effect analysis is performed on the temperature field reconstruction results of continuous time windows to extract the axial temperature gradient, radial heat exchange efficiency and eddy current intensity coefficient.

6. The method according to claim 1, characterized in that, The thermostat response performance evaluation index is generated based on the deviation between the actual displacement trajectory of the valve core and the target stroke control command, combined with the cooling medium flow rate change rate. This index includes: Calculate the following delay time and overshoot of the actual displacement trajectory of the valve core relative to the target stroke control command; Analyze the deviation between the actual response rate and the expected rate of the cooling medium flow rate change. By integrating the following delay time, overshoot, and deviation, a dynamic response error coefficient is generated. Calculate the response baseline correction factor based on the ambient temperature and engine oil viscosity parameters in the historical control data; The dynamic response error coefficient is multiplied by the response benchmark correction factor to generate the thermostat response performance evaluation index.

7. The method according to claim 6, characterized in that, The step of dynamically correcting the thermal inertia delay parameter in the thermostat dynamic response model based on the thermostat response performance evaluation index includes: Establish an inverse mapping relationship between the thermal inertia delay parameter and the thermostat response performance evaluation index; The compensation amount for the thermal inertia delay parameter is calculated using the gradient descent algorithm; A first compensation weight is applied to the medium thermal conduction time constant in the thermal inertia delay parameter; A second compensation weight is applied to the metal heat storage coefficient, and the second compensation weight is related to the specific heat capacity of the engine block material; The thermal inertia delay parameter is updated based on the weighted compensation result.

8. The method according to claim 7, characterized in that, The step of updating the target stroke control command based on the corrected thermal inertia delay parameter includes: The updated thermal inertia delay parameters are fed back to the thermostat dynamic response model; Recalculate the hydraulic actuation response function in the valve core displacement field equation; The engine operating condition intent vector, the updated hydraulic actuation response function, and the cooling medium flow resistance characteristic function are integrated through a multi-dimensional control vector fusion algorithm. Generate updated target travel control commands that include position feedforward compensation and flow feedback correction.

9. The method according to claim 4, characterized in that, The method further includes: Monitor sudden environmental temperature events and changes in the composition of the cooling medium; When a sudden change in ambient temperature is detected, a thermal stress compensation mechanism is triggered to perform thermal shock correction on the cooling medium temperature fluctuation matrix. When a change in the composition of the cooling medium is detected, the viscosity-temperature correction factor in the flow resistance characteristic function of the cooling medium is adjusted. The corrected cooling medium temperature fluctuation matrix and flow resistance characteristic function are input into the stroke prediction model to recalculate the target stroke control command.

10. The method according to claim 9, characterized in that, The method further includes: Collect deviation data between the actual temperature adjustment curve and the expected temperature curve within the preset verification period; The travel prediction model is trained using residual backpropagation based on the deviation data. Update the weight parameters and convolution kernel coefficients of the feature extraction layer in the trip prediction model; The updated travel prediction model will be applied to subsequent control cycles.

Images